Skip to main content

tv   House Intelligence Hears from Twitter Facebook and Google  CSPAN  November 7, 2017 1:55am-4:39am EST

1:55 am
officials from google and twitter were on capitol hill. the witnesses discussed steps their companies were taking to address the distribution of false advertisements and the creation of fake accounts. this house intelligence hearing is just under three hours. good afternoon. the hearing will come to order.
1:56 am
i'll like to offer up a brief prayer for what happened in new york. we come to you today humble, seeking your wisdom, guidance, knowledge as we try to lead this great people. lord, we have circumstances in new york where families are in deep pain and deep sorrow and tragedies in california and florida and others, puerto rico where families are struggling. ask your devine guidance and comfort to first responders and others. we ask, lord, that we're worthy of your praise and blessing and that we conduct ourselves this afternoon in ways that we'll honor you. -- ask in jesus name, amen. appearing before the committee will be mr. collin stretch, general counsel of facebook,
1:57 am
acting general counsel of twitter and kent walker, senior vice president and senior counsel of google. we are in open session. this will only unclassified matters. we appreciate the public and media interest in this committee's work. we expect proper decorum at all times. we ask the witnesses to stand and raise their right hand. do you solemnly swear or affirm the testimony you give to this committee will be the truth, the whole truth and nothing but the truth, so help you god. thank you. you may be seated. i'll recognize myself for five minutes. thank you for being here today to discuss a very important topic and that's russia's use of your social media platforms during the 2016 election. as you know, this committee has been investigating russia's involvement in the 2016 election since the beginning of this year. as part of that investigation, we're examining the role that social media companies played in
1:58 am
the disseminating maligned content produced and paid for by russian actors including the russian propaganda arm, the internet research agency. it is no secret russian actors used your social media platforms after and during the election cycle to communicate discord among our citizens. such tactics are not new or novel but the manner which they can be employed using social media is unique. for example, let's examine some of the facebook advertisements that were identified as being created by russian actors. these images were provided to the committee in advance of today's hearing and represent a small sample of some of the images that appear on computers and mobile devices between 2015 and 2017. these exhibits are not selected
1:59 am
for political gain but to provide a clear example of what we seek to discuss. these exhibits some of the most viewed russian created facebook advertisements in 2015 to 2017. exhibit 1, if you'll put that up. is an ad entitled being patriotic. it was created on june 23rd, 2015, and received approximately 530,000 ad impressions and 72,000 ad clicks. it costs approximately 330,000 rubles. exhibit 2 was created on december 10th, 2015. and received approximately 531,000 ad impressioned and 30,000 ad clicks. it cost approximately $2,100 at today's exchange rate. exhibit 3 is entitled south united. it was created on october the 14th, 2016. and received approximately 40,000 ad clicks. this cost approximately $1,300 at today's exchange rate.
2:00 am
exhibit 4 is an ad entitled back the badge. it was created on october 19, 2016, and received approximately 1.3 million ad impressions and 73,000 ad clicks. this cost approximately $1,900 in today's exchange rate. finally exhibit 5 is an ad entitled woke blacks. it was created on december 1st and received approximately -- cost approximately $1,000 in today's exchange rates. i ask unanimous consent that all exhibits and material offered in today's hearing be entered in the record. without objection so ordered. all three copies have a public responsibility to ensure that the content carried on your platform is not produced by foreign adversaries seeking to harm our society and democratic process. i submit this is not easy in a democracy where free speech is
2:01 am
guarded by our constitution. americans must always be free to pick and choose which stories to read, click or retweet. we must not let technology foreign enemies to spread this information with the intent to divide us. i thank you for your efforts your companies have recently made to address the harmful russian influence on your platforms and the transparency which you have made those changes. i hope today's hearing will help the public understand the extent to which russian actors use your platforms during the 2016 election. we also help your testimony will shed light on the relative impact of this hostile influence compared as to other legitimate messaging campaigns during the same period. and how your companies distinguish between malign activities and free speech. to include whether these differences filter content and russia democracies as opposed to china and russia. we expect you will address your company's specific plans moving forward to help ensure such
2:02 am
activities do not occur again on your platforms. i look forward to a productive hearing. i will recognize the ranking member mr. schiff for five minutes with any opening comments. >> thank you, mr. chairman. in march of this year our committee had its first open hearing and then fbi director comey revealed that he'd opened a counter intelligence investigation involving trump associates and the russians. then we knew next to nothing about the russian's use of social media to attack hillary clinton. the technology companies themselves only recently have identified the reach of that facet of the kremlin's campaign. today we will see a sample of those ads and ask the social media companies what they know about the extent of russian extent of social media, what they intend to do about it to protect our country from this malign influence in the five. i posed the question whether the trump campaign colluded with russia in any aspect of its influence operations. in essence did the russians offer to help the campaign and did the campaign accept.
2:03 am
and if the trump campaign did accept explicitly or implicitly, what did the russians do to make good on that understanding? we now know as a result of the guilty plea by trump campaign foreign policy adviser gorge papadopoulos that the russians approached the trump campaign as early as april of 2016 to inform them they were in possession of dirt on hillary clinton in the form of thousands of stolen e-mails. this timing is significant because it means that the trump campaign was informed of russia's involvement with the stolen e-mails even before our country was aware of it. but mr. papadopoulos was not the only trump campaign figure the russians approached. nor would his lies to federal agents be the last example of trump associates making false statements about their interactions with the russians. we now know that the uppermost levels of the trump campaign were also informed that the russians had dirt on clinton and that it was offered to the campaign in what was described as part of the putin's
2:04 am
government effort to help mr. trump. that offer appears to have been accepted when the president's son said that he would love the assistance and suggested that the best timing would be in the late summer. and in late summer the russians would begin dumping dirt on hillary clinton. the president and his son would later deceptively claim the meeting with the russians in trump tower was about adoptions. it is not clear from mr. papadopoulos's plea whether the russians communicated the -- would not involve the direct provision of the stolen e-mails to the campaign, but their publication through wikileaks and moscow's own cutouts. what is clear is this. the kremlin repeatedly told the campaign it had dirt on clinton and offered to help it and one top official, the president's own son accepted. apart from publishing stolen e-mails, the russians always used social media to assist the
2:05 am
trump campaign. whether the russians and the campaign coordinated these efforts we do not yet know but true the russians mounted what could be described as an independent expenditure campaign on mr. trump's behalf. russian ads on twitter promoted stories about hillary clinton's allegedly poor health. discord in the u.s. by inflaming passions on a range of divisive issues. the russians did so by weaving together fake accounts, pages and communities to push politicized content and videos and to mobilize real americans to sign online petitions and join rallies and protests. they also bought ads like these, the first ad behind me, black matters was brought to us from our friends in st. petersburg. the second ad garnered over 135,000 facebook followers. the russians bought sufficient
2:06 am
ad space for it to appear almost 145,000 times across facebook accounts, that it expressed an interest in donald trump stopping illegal immigration, conservativism, federal states of america or the republican party. russia exploited real vulnerabilities that exist across online platforms and we must identify, expose and defend yourselves against covert operations in the future. the companies here today must play central role while preventing cyberspace from being misused by our adversaries. i thank you, mr. chairman and yield back. >> we will now turn to our witnesses. we have your opening statements for the record. written prepared statements. do any of you care to make an
2:07 am
oral statement to the committee at this time? if so, you'll have five minutes. >> chairman, ranking member shift and members of the committee, twitter understands the importance of the committee's inquiry into russia's interference in the 2016 election and we appreciate the opportunity to be here today. the events underlying this hearing have been deeply concerning to our company and the broader twitter community. we are committed to providing a service that fosters and facilitates free and open democratic debate and that promotes positive change in the world. we are troubled by the reports that the power of twitter was used for the purpose of influencing the presidential election and undermining public faith in the contract process. the abuse of our platform to obtain manipulation for election system a new challenge and one that we are determined to meet. today we intend to show the committee how serious we are
2:08 am
about addressing this new threat by explaining the work we are doing to understand what happened and to ensure that it does not happen again. at the time of the 2016 election, we observed and acted on instances of automated and malicious activity. as we learned more about the scope of the broader problems we resolve to strengthen our systems going forward. elections continue all the time. so our first priority was to do all we could to block and remove malicious activity from interfering with your users experience. we created dedicated teams within twitter to enhance the quality of information our users see and to block malicious activity whenever and wherever we find it. those teams continue to work every day to ensure twitter remains a safe, open, transparent and positive platform. we have also launched a retrospect tiff review to find russian efforts to influence the 2016 election through automation, coordinated activity and advertising. while that review is still under way, we have made the decision to share what we know today. in the interest of transparency
2:09 am
and out of appreciation for the urgency of this matter. we recognize that our findings may be supplemented, discover more facts and gain a greater understanding of these events. my testimony details the methodology and current findings of the review and detail. we studied tweets from the period september 1st through november 15th 2016. we did find automated and coordinated activity of interest. we determined that the number of accounts we could link to russia and that were tweeting election related content was comparatively small, around 1 100th of a%. 1/3 of election related tweets came from russian linked automated accounts. we did, however, observe instances where russian linked activity was more pronounced. and we have uncovered more accounts linked to the russian
2:10 am
based internet research agency as a result of our review. we also determined advertising by russia today at seven small accounts violated the policies in effect at the time or that have since been implemented. we have banned all of those users as advertisers and we will donate the revenue to academic research into the use of twitter during the election and for civil engagement. last week we announced industry leading changes to our advertising policy. we are also enhancing our safety policies. sharpening our tools for stopping malicious activity and increasing transparency to promote public understanding of all of these areas. our work on these challenges will continue for as long as malicious actors seek to abuse our system and we will need to evolve to stay ahead of new
2:11 am
tactics. we have heard the concerns about russian actors use of twitter to disrupt the 2016 election and about our commitment to addressing this issue. twitter believes that any activity of that kind, regardless of magnitude, is unacceptable. and we will agree to do as much as we can to do better to prevent it. we hope that our appearance today and the description of the work we have undertaken demonstrates our commitment to working with you, our industry partners and other stake holders to ensure that the experience of 2016 never happens again. cooperation to combat this challenge is essential. we cannot defeat this threat alone. as with most technology based threats the best approach is to combine information and ideas to increase our understanding. working with the broader community, we will continue to learn, to test, to share, and to improve so that our product remains effective and safe. i look forward to answering your questions. >> thank you, sir. mr. stretch. >> chairman, ranking member schiff and distinguished members of the committee, thank you for this opportunity to appear before you today. my name is collin stretch. since july, 2013, i've served as
2:12 am
the general counsel of facebook. we appreciate your hard work to investigate russian interference in the 2016 election. i'd like to start by echoing the comments of the chairman earlier regarding the events in new york yesterday. we extend our deepest condolences to the victims of this horrific attack and to their families. we're doing everything we can to assist law enforcement with its investigation. at facebook our mission is to create technology that gives people the power to build community and bring the world closer together. we are proud that each of you uses facebook to connect with your constituents and we understand that the people you represent expect authentic experiences when they come to our platform to share and to connect.
2:13 am
we also believe we have an important role to play in the democratic process and a responsibility to protect it on our platform. that's why we take what's happened on facebook so seriously. the foreign interference we saw during the 2016 election is reprehensible. that foreign actors hiding behind fake accounts abused our platform and other internet services to try to sew division and discord and try to undermine our election process is contrary to our values and everything we stand for. our goal at facebook is to bring people closer together. these foreign actors sought to drive people apart. in our investigation, which continues to this day, we found that these actors used fake accounts to place ads on facebook and instagram that reached millions of americans over a two year period. and that those ads were used to promote pages which in turn posted more content.
2:14 am
people shared these posts, spreading them still further. many of these ads and posts are inflammatory. some of down right offensive. we know that much of this content is particularly hurtful to members of the facebook community that engaged with this content believing it was authentic. people should believe content on facebook is authentic. and should not have to worry that they are being exploited in a cynical effort to prey on painful fault lines in order to assume discourse in the country. the ads were a very small fraction of the overall content, but any amount is too much. all of these accounts and pages violated our policies and we removed them. going forward we are making significant investments. we're hiring more ad reviews, doubling or more our security engineering efforts, putting in place tighter ad content restrictions. launching new tools to improve a transparency.
2:15 am
we're building artificial legs to locate bad actors. we're working closely with industries to share information on how to identify threats so we can all respond faster and more effectively. and we're expanding our efforts to work more closely with law enforcement. we know bad actors aren't going to stop their efforts. we know we'll all have to keep learning and improving to stay ahead of them. we also know we can't do this alone. that's why i want to thank you for this investigation. we look forward to the conclusions you will share with the american public and i look forward to your questions. >> walker, do you have a statement? >> i do. >> thank you. >> chairman conaway, ranking member shift, thank you for the opportunity to be with you. my name is kent walker. at google, responsible for trust and safety. i've worked at the intersection of law, technology and security
2:16 am
for over 25 years including a stint early in my career as an assistant u.s. attorney at the u.s. department of justice specialized in technology crimes. let me start my conversation this afternoon by adding my acknowledgement to the victims and the families of the awful attack that happened yesterday in new york city. as a new york city employer, we know how strong and how tough new yorkers and we look forward to doing whatever we can. turning to the issues before the committee today, google believes we have the responsibility to prevent the misuse of our platforms and we take that responsibility very seriously. google was founded with the mission of organizing the world's information. the abuse of our platform is -- we are committed to working with the congress, law enforcement, others in our industry, to -- we recognize the importance of this committee's mandate and we
2:17 am
welcome the opportunity to share information and talk about solutions. of course disinformation and propaganda campaigns aren't new and have involved different types of media and publications over the years. we have seen attempts to interfere with our online platforms. we built industry leading security systems and put them into our consumer products as well. back in 2007 we launched the first version of our safe browsing tool which helps protect users from fishing and other attacks. if and when we suspect they're exposed to -- last month we launched our advanced protection program which helps those protest those at greatest risk
2:18 am
of attack. we face motivated and resourceful attackers and we are evolving our tools to stay ahead of ever changing threats. our tools don't just protect our physical and network security. they also detect and prevent attempts to manipulate our systems. google news for example we now use fact check labels to help users spot fake news. for google search we updated our quality guidelines and evaluations to help surface more authoritative content. we've updated advertising guidelines to prohibit ads that misrepresent themselves and on youtube we employ a sophisticated spam and security breach detection system to detect behavior and check people trying to inflate view counts or number of subscribers. as threats evolve we will continue to adapt to avoid new attempts to misuse our platform. with respect to the committee's work to the 2016 election, we have looked across our product to see where the government back entities in order to interfere with the u.s. elections. while we did find some deceptive activity, that activity appears to have been limited.
2:19 am
any activity is more than we'd like to see. we've provided the information to the committee, and will continue to cooperate with the committee's investigation. going forward we will continue to expand our use of cutting edge technology and continue working with governments to ensure our platforms aren't abused. we'll also make political advertising more transparent, easier to understand and more secure. in 2018 we will release a transparency report and how much money is being spent. we'll pair that with a database available for public research of election and content from across our ads products. we're ailes going to make it easier for users to understand who bought the election ads they've seen.
2:20 am
going forward users will be able to find the name of any advertiser running an election ad on search, youtube and the google display network from a icon on the ad itself. we will continue enhancing our safeguards to ensure we permit only u.s. nationals to buy u.s. election ads. moving forward we'll go further. verifying the identity of anyone who wants to run an election ad or use our political interest based tools and confirming that person is committed to run that ad. we certainly can't do this alone. we'll continue to work with other companies to protect our digital ecosystem. even as we take our own steps, we'll remain open to working legislation that promotes transparency. our commitment reaches to --
2:21 am
we've offered in person briefings and introduced a suite of digital tools designed to help election websites and political campaigns protect themselves from fishing, unauthorized account access and digital attacks. we are also increasing our support for the defending digital democracy project. let me conclude by recognizing the importance of the work of this committee. our users, advertisers and create ors must be able to trust in their safety and security. we share of goal of identifying bad actors who attempted to interfere with our systems and abuse the electrical process. we look forward to continues co praying to provide access to cools and help citizens express themselves while avoiding abuses that undercut the integrity of elections. thank you again for the opportunity to tell you about our ongoing efforts.
2:22 am
we look forward to continuing our work on these issues and i'm happy to answer any questions you might have. >> thank you are figure here. your testimony i for one recognize myself for five minutes. thank you for what you're doing and the efforts. what i heard mr. walker just say by previous conversation with the other two companies, you're investing significant corporate resources and putting weight behind the comments and the commitments you've made today. mr. stretch, the numbers on our -- the five exhibits that we showed and certainly the two that he showed, can you tell us what the difference is between an impression and an ad click is and then some brief comment as to context as to those look stunningly impressive just on their face, but against a broader backdrop, can you help us put that in context. >> an impression, congressman, is content that is in view for a user. it doesn't necessarily mean that a user stopped and viewed. if you think about how you use your phone and open up an app and scroll through it, anything in there would be an impression. a click means engagement with the ad. so with these ads the click may have been to like the content. >> so one of the questions is
2:23 am
how much influence did these ads and information, misinformation have? are there metrics that you use as part of your evaluation of your normal business model as to, you know, if you're trying to help a company develop an ad program? you say we're going to show you -- we think your add will have this kind of impact? are there tools like that you can use to evaluate what impact these ads might or might not have had with respect to the opinions that americans were forming as to who to vote for in november '16? any of the three? >> we have tools certainly to help advertisers measure their return on investment. those are typically for more larger advertisers to help them understand different campaigns. for campaigns like we saw from the accounts we've subsequently linked to the internet research
2:24 am
agency, they're typically or they were intended to drive followership of the pages, so getting people to like the page, for example. and there the return on investment really is clear from how many people liked the page. >> so would you consider the return on the rubles invested they get a return on the money make sense, or were they under or over expectations? >> congressman, i can't say what their expectations were. i do think it's clear that they were able to drive a significant following. >> they were or were not? >> they were able to drive a relatively significant following for a relatively small amount of money. it's why this activity appears to pernicious. it was undertaken i think by people who understand social media.
2:25 am
these people were not amateurs. i think underscores the threat we're facing and why we're so focused on addressing it going forward. >> in looking at the ad, based on what happened in '16, no one looking at that could distinguish that from, say, a left wing, right wing group who might have been trying to pitch that same message with those images. would there have been a way a user could have distinguished that was a foreign actor versus someone here in the united states that might have a horrible opinion but they wanted to use that platform? could they tell? >> on the face of the content, i think it will have been difficult to do so. >> all right. are you doing something -- we'll get to the other two folks, maybe to the other conversation, but is facebook doing something looking at the '18 election that
2:26 am
would help users see who that is and that kind of thing? >> we're taking a number of steps with respect to elections going forward, congressman. we are investing in our security efforts to make sure we're better policing the authenticity of the site. we never want to see that content on the site in the first place, because it is such an effort to -- >> even if it's someone american, it would violate your standards? >> right. it's an excellent question. we believe that when people show up to facebook as their authentic selves, they have the opportunity and should have the opportunity to speak on important social issues like the ones that are discussed in these ads. the problem with these ads and they should not have run on the site is people weren't showing up as their authentic selves and it undermines the trust that is so important to our platform. >> do you think you'll have tools available before the '18 election that would allow someone to look through the face value of the ad to see who did it and how much and that kind of
2:27 am
thing? >> if i may, i'd like to make two points in response. one, we are and have already incorporated the learnings from what we've seen from this sort of behavior into our automated tools so that our automated tools are better able to detect and rid the site of these masquerading accounts. the second point to your question of disclosure, absolutely. with respect to political ad is in particular, we want to give advertisers an opportunity to make clear who's behind the ad. and where we see political ads that don't come with that disclosure, that will be a very strong signal for us to require information and documentation to make sure that people who are running political ads in connection with the u.s. federal election are authorized to do so. >> all right. thank you, gentlemen. i turn to my colleague adam shift. five minutes. >> thank you, mr. chairman. the january 2017 intelligence community concluded russian -- that blends covert intelligence operations such as cyber activity with overt efforts by the russian government agencies,
2:28 am
state funded media, third appeared intermediaries or trolls. we now have a much better sense of how that manifests itself. on facebook we learned of 470 fake accounts tied to the kremlin linked internet research agency or troll farm. from these accounts more than 80,000 individual pieces of organic content were produced such as posts to which at least 126 million americans were exposed. roughly 3,400 paid ads were purchased by the troll farm over the period from june, 2015, to august, 2017. ads which over 11 million americans saw during the campaign season. on twitter roughly 2,700 human linked twitter users connected to the kremlin troll farm who tweeted 1.5 million times and 36,000 russia linked bots were also found. these tweeted 1.5 million times which accounted for nearly 300 million views. with respect to google and youtube, roughly 1,100 kremlin linked videos were posted to
2:29 am
youtube for about 300,000 views. the question i'd like to ask you all, really a couple questions, one that is very specific. one that is a broader significance. part of what made the russian social media campaign successful, part of mr. stretch as you point out why the russians were sophisticated media is they understood the algorithms that you use tend to accentuate content that is fear based or anger based. that helps it pick up an audience and go viral and be amplified. this is an issue of not just foreign -- which the algorithms are designed to attract attention and keep our eyes focused on the platform for advertising purposes may also have the unintended consequence of widening divisions among our society, polarizing people. what ends up percolating the top of our feeds tends to be things we were looking for or things
2:30 am
the algorithms will capture our attention. my question is what corporate obligation, societal obligation do you think your companies have using both of these issues? the foreign manipulation of your platforms, but also more broadly the fact that algorithms designed to attract our attention may also have the unintended consequence but very real consequence of pitting american against american in a way that the russians so capably manipulated? so if you can each address that question. the second question is do you have the historic data such that you would be able to analyze the trump campaign advertising and its campaigns, organic content with that produced by the russian social media farms? and analyze whether in its targeting or its audience there was any sophistication in that overlap. >> i'll start. we obviously take both of these
2:31 am
issues very seriously. i think you'll hear throughout our conversation today that our focus, while we do look at content and we have rules that talk about content, you know, when it turns violent and twitter turns violent, we have the greatest protection -- when we look at behavior and the information we see behind twitter accounts. we talk about things like automated malicious accounts because what we've seen, especially in this investigation, is that these malicious actors need -- they need ears. they need eyes. they need to be able to reach an audience. and the way they get that audience without being able to grow organically is to use automated activity on the platform. and that's where we're focused. that's where we've gotten a lot better. over the last year we've improved our ability to
2:32 am
challenge accounts. we're challenging four million accounts every week to determine if they're real. we take down and block 450,000 suspicious log-ins every day. we're making a concerted effort to stop this type of activity, to give a voice to the people who are trying to abuse our system. as to the ira and the russian based troll farms that we've been able to identify to date, we do have that information and
2:33 am
can share it with you. >> mr. stretch and mr. walker. >> gentleman, being respectful, i'd ask you to be brief. >> i will be brief. yes, we do have an obligation to prevent foreign interference in the election. we take that obligation seriously. there are more details in my written testimony how we're attempting to discharge that. with respect to the algorithm, our goal is to provide the most relevant information. it's primarily driven by friends and families. that's the core use of facebook. we want you to come to facebook and see information that's important to you. typically that's the information that's important to your friends and to your family. in a political election season, often times what's important to your friends and family are challenging, provocative social issues. so you will see that. our responsibility is to make sure that when you see that
2:34 am
content, it's authentic. so that you can trust the dialogue that's occurring on the platform. then to your last question, we have not seen overlap in the targeting that was relatively rudimentary used in the ira ads that we've disclosed any other advertiser that's been operative on the site including the trump campaign. >> just briefly, the accuracy and integrity of our results is the north star of our work at google. that goes to questions of fake news as well as efforts to interfere with the electoral process in the united states or anywhere around the world. i'd be happy to detail the steps we have taken and continue to take to safeguard our users from that kind of abuse and interference. with regard to the targeting of ads, the use of platforms for advertising is relatively limited. about $4,700 and generally not microtargeted but we'd be happy to answer questions. >> time expired.
2:35 am
five minutes. >> thank you, mr. chairman. thank you all for being here. social media flat forms have the responsibility before striking balance and removing false information and providing freedom of speech. can you give us detail about how each of your companies plan to target false news while protecting the robust political discourse? >> let me take that because that was the next stage to my answer to mr. schiff's question. we are taking a number of different steps beyond advertising to focus on fake nice. we are working to improve our algorithms, to provide additional guidance and training to the raiders who provide quality feedback and look at a wider variety of signals to improve the ranking of authentic and genuine news and demote sites that we feel are deceptive or misleading. we're also making broader use of fact check labels. working with third parties for both google search and google news. when it comes to advertising,
2:36 am
we've taken steps to disallow advertising on sites that misrepresent their nature or purpose and to add to our policies around or against hate speech, encitement of violence special the like. >> first we find that most false news is financially motivated. we're making efforts to disrupt the financial incentives that we think will make a dent. second, we're looking to stop the spread of it so when we have information that's been disputed by independent fact checkers, we limit the distribution and alert users who are attempting to share it that it has been disputed. we're engaged in education efforts to help particularly around the world users approach some of the content they see with a more discerning eye. >> we're tackling this challenge in a few ways. the way it was characterized is
2:37 am
correct. it's a balance between free speech and what's real and what's false. we often see there's a lot of activity on the platform to correct false narratives. one of things is the text to vote tweets which we took off our platform. the number of tweets that were counter acting that as false and telling people not to believe that was, like, between eight and ten times what we saw on the actual tweets. but we're working on the behavior. that's where we're focused on right now for things like terrorism and child sexual exploitation. we're trying to figure out how we can use those learnings to stop the amplification of false news. it's a definite balance. we also have work we're done just like my peers around ads
2:38 am
transparency that i think is going to help educate the consumer about who's paying for an ad, what they're targeting, what they're after. especially around election ads, who's paying for it, how much they're spending. we're also working with third parties around the world who are helping us think through the thing we are trying to employ, how to tackle the issue and how they will debate free speech on our platform. we're working hard, but it's a challenge. >> so i know all of you have said you've committed significant corporate resources into this, but i think a prime question that i think certainly the committee has and i think the entire country has is what assurances can you give us that foreign malicious activity in the 2018 elections and beyond are going to be mitigated? >> i can assure you we are focused on it and we are improving. we see really opportunities for
2:39 am
improvement in three categories. first, we have to be better technically. we have learned a lot from the 2016 election cycle and from the political trolling behavior. we've seen really worldwide in the last year or so. and we've incorporated that learning into our automated systems and are seeing results. the second area where i think we have room to improve is in industries cooperation. we think there's a real good model for this in terms of how we have shared expertise and threat information in other areas of abuse on the platform. we're looking forward to standing that up in the area as well. third and finally, we think a constructive dialogue with law enforcement authorities where,
2:40 am
again, we're sharing information with respect to specific threat actors as well as expertise about how they're operating will be mutually beneficial and put us in a much stronger position as we head into next year's elections. >> since my time is expired, being respectful of the other members, if you could maybe get those answers back to us to the committee that we can refer to. thank you. i yield back. >> mr. hines, five minutes. >> i'd like to use my short time to explore russia's use of twitter. i'll be directing my questions to you. first, in a few short words, can you please explain to us the difference between a bot and a troll. >> the way we think about that internally is a bot is an automated account so it's an account where the machine is largely responsible for the actions, setting it up, tweeting, retweeting. >> so it's fully automatic? >> yes. typically that's the behavior we see. we do see combination. troll farms are a new challenge for us and a bigger challenge we're going to try to tackle in a few ways. when we think of trolls we think of a real human behind the
2:41 am
account. oftentimes coordinated with others or coordinated with the few or many accounts. >> so with the troll, it's a real human but necessarily a real human who we know who that individual is. >> typically that's the behavior we see. >> and if i understand this correctly, the russians took advantage by creates false accounts that were trolls and deploying many bots that were able to retweet into the thousands they thought convenient to their cause. >> that's what we discovered. >> here's kind of the key issue. it's essentially to twitter that there is not a requirement that the person disclose their true identity on their platform. >> that's correct. >> so these are important points, because how russia used twitter is necessarily different from how they used other platforms like facebook and youtube where there isn't the same anonymity. in particular the use of anonymity means a kremlin linked user in russia or ukraine could tweet and share content without
2:42 am
anyone knowing who they were. they could pretend to be a person and the everyday user has no way of knowing who they are, right? >> that's correct. we have a number of signals behind the account we can share with law enforcement when necessary and we do verify a number of individual accounts, both corporate accounts and individual accounts. >> so let me give an example. the board behind me shows a few of over the 2,700 twitter users you have discovered that are connected to the kremlin's internet research agency. looking at the names, there would be no way for the user to know seattle post, that the content was in fact generated or retweeted by a russian entity. >> the real seattle post would
2:43 am
be verified presumably on the platform, but they wouldn't know just by looking at that user name. >> should political content created on one hand by bots or by any other form of artificial intelligence, should that be labelled as such? and if that content is generated by a foreign person. should it be labelled as such? >> so on to your first point on automation. we're not only trying to -- we don't try to label it, we try to remove it. we see automated retweets, we're removing those actors from the platform. because of the information we have behind the scenes, we can connect those accounts often times. because of the information we have behind the scenes, we can connect those accounts often times. we're not just removing the one, we're removing the collective. >> what go think your success in removing bots is? >> we're getting better. we think we've gotten twice as good in the last year. we're challenging 4 million accounts a week. >> give me a sense of percentage overall. put that in context for me. >> context of what?
2:44 am
>> i don't know what twice as good means unless you tell me. >> last year we were challenging 2 million accounts a week based on the technology. as we learn from activity of automated accounts and their patterns and getting better at detecting them, we get better, because those old techniques don't die. and the new techniques we learn from or get ahead of 2 million accounts last year this time. 4 million this time 450,000 suspicious log ins a day, we're taking them off the platform before they even retweet. >> part of the power of twitter russia sought to abuse was to make real users share russia's propaganda to a wider audience. the board behind me tweets purporting to be the tennessee gop. this was reportedly shared by trump campaign officials, donald trump
2:45 am
jr. followed the account until it was shut down in august. in sharing this content, trump's campaign knowingly or unknowingly, helped legitimize and spread russia disinformation. i'm a big fan and big user and i respect your desire to keep an open discourse. i hope you remaining committed to uncovering this kind of meddling in the core of our democratic process. >> we're very committed. >> when you say you don't necessarily label something that you come to find out is false, that you try to remove it. i just have to be honest with you, i don't personally use twitter, so no offense, but what -- how hard would it be for them to take down seattle post and do seattle post one after you take it down? >> i mean rather than letting people know by the way, this is a russian produced propaganda ad
2:46 am
or this is a foreign-produced news story, so take that for what it's worth. just trying to remove them as they go, aren't you chasing your own tail all day long? >> it may appear to be a game of whack-a-mole. there's a lot of signal we see behind the public facing site. we see where people log in from. the devices they're using, the phone numbers they're using. their ip address. we're able to see and stop someone who's a bad actor. we're able to use signals to maintain the database of bad actor locations and other signals and stop the act creation. we get better every day, that's why our technology and results are getting better all the time at stopping these things. but we're using the signal and behaviors behind the scenes, to stop them before they create
2:47 am
another account like you said so we're not playing that game. >> i appreciate that, and appreciate what you said about blocking malicious activity. mr. stretch, you know, trying to figure out policing authentic cells. these are new terms unfortunately becoming our new normal in this country. the problem i have, i don't know how successful you're being so far, to this day, we see news stories that come out that we come to learn in short order, as recently as a lot of the nfl sneeling scandal was perpetrated abroad to try to pit ourselves against each other. taking both sides and just throwing it out there. it's as recent as that. my question to you is, i don't know if i have an opinion on this or not, but we've talked about on both sides of the
2:48 am
aisle, do we have a role in this body in assisting you in trying to figure out for foreign entities, not american. not american journalists, certainly, i might say something that's completely opposite of what somebody on the other side of the aisle, in two different networks, both believing it's true. you can't police who's right or who's wrong. that would be a violation of my free speech. with regard to foreign entities trying to infiltrate and create propaganda against our citizenry, does the united states congress have a role in assisting you. if we do, what would that be in your opinion in alerting my constituents in okeechobee florida that this piece of news you're reading, not just a political ad, but a news story on facebook is not true. and i know that because there's a label or disclaimer or something on there that shows, by the way, what you're reading was produced in a foreign country. do we have a role in that?
2:49 am
and if so, what is that. and how can we make sure we're not violating people's constitutional rights by getting involved in that? >> so the challenge you identified is an acute one. we don't want to put ourselves in the position of being the arbiter of truth, we don't think that's a tenable position for any company or industry to bear. we think it's inconsistent with protection of personal protection that's so foundational to this country. we are taking a number of measures to ensure again the authenticity and the trust is present on the platform including labeling stories that have been disputed as false. where we see a role for government in assisting in this
2:50 am
effort is to ensure that we are all sharing information about the techniques and threat actors that we need to be alert to. and monitoring on the platform and disrupting when they engage in this sort of activity that chairman and the ranking member surfaced earlier. that's where we feel like there's really the opportunity to come together not just as an industry but as a country to work on this problem together. >> anybody else? >> i second that, i'd just add any additional leads the government has that they can provide would be helpful. >> i yield back. >> thank you, mr. chairman. gentleman, i think there's no doubt that russia tried to use your platforms to weaponize and meddle in our elections. i think that it has risen to the level of a national security issue. and, therefore, while you're self-policing yourself, and
2:51 am
that's great, i really do believe that we have an obligation to the american people to do more than that. my line of question is really to facebook. i understand that there are paid ads, political ads as well as what you guys call organic postings that are not paid for. an example of the paid ad would be the one here from a facebook page called being patriotic that urges those that follow it to go to a rally. it was a made up rally. my constituents the people i talk to at town hall meetings can't understand how you go about vetting your content and the person who is your user. comes to finds out, being patriotic is one of those troll farms that's russian. and it was paid for, the
2:52 am
political ad was paid for by rubles. isn't that a red flag? how could that happen? how do you vet your content and your users? >> you've identified two very important areas of investment for us and opportunities for us to be better. first, on the ad side, we are tightening our policies. all ads on facebook go through a combination of automated and manual review. and we're tightening our tools to make sure that ads that are on potentially socially divisive topics get heightened review. and we are -- >> could i -- i only have a short period of time. i want to know, who are your vetters? is it content analysts? if so, who are these people? are they experts? are they average yo-joes?
2:53 am
how diverse are these voters? the reason i ask is because if you look at your organic postings. an example of that is blacktivist, also turns out to be a fake facebook page done by one of those russian troll farms. and that is trying to insight racial animosity, my point is this, with all due respect, i know that you all are good corporate citizens and you consider yourself to be such. i think it is paramount to our national security to have more transparency and more accountability on all of your platforms. i know i'm talking to facebook, do know i am really broadly talking to everyone. i want to know what you see as your responsibility to make sure that you are not -- that you are actually vetting the content. line because of free speech.
2:54 am
and that is a paramount foundation of our democracy. who are your vetters? are they a diverse group of people? >> the vetters, the people who work on ad review, they are around the globe, so we have a number of languages and regions that we cover, so we have people around the globe. like every aspect of our workforce, we are committed to building a workforce as diverse as the community we serve. >> with all due respect i have to stop you there. i'm a member of the congressional black caucus, i know that just last week, several of my colleagues went to facebook to meet with your executives to talk about your diversity of initiatives. i don't know if you know exactly how many racially diverse workforce you have, what the percentage is. i can tell you if you don't know. it's very low. the reality is, facebook over overall racial ethnicity with
2:55 am
respect to black employees workforce-wide is 8.8%. with respect to your leadership, it's 2.3%. and you're saying i should trust that your vetters that are going to be vetting this kind of information will be a diverse workforce? >> congressman, i appreciate the feedback, and we value the input from the meetings last week. it is, what you should be confident of is that we understand the importance of diversity. we are committed -- >> with all due respect, i have nine seconds. what i want to say is this. i submit to you that your efforts have to be more than just about finding malicious and deceptive activity, you have a responsibility, all of you have a responsibility to make sure that we are not adding to the problem by not being as rigorous and as aggressive as we can in terms of vetting the content and in terms of making sure that we
2:56 am
are being really dynamic in doing that, i also want to say that i think it's ridiculous that a foreign entity can buy a political ad with rubles but can't give a political contribution to me, a russian person can't give a political contribution. there seems to be some legislation that needs to be had here, is all i'm saying. >> thank you, the lady's time has expired. mr. turner, five minutes. >> thank you, gentlemen. there have been controversies before about content on each of your platforms, from democrat anthony weiner on twitter with issues on facebook of isis and terrorism recruitment, and radicalization, to concerns with google and biases of search engines. there's a concern about algorithms and data and misrepresentation of material, how individuals are targeted. and that really is my question
2:57 am
that i have to you, because the last questioner made an excellent point, it's not just the postings that occurred because you talk of yourself as being communications vehicles. many times you're not looked at as neutral communications vehicles. not just content and users, the posting also ads -- the organizations were paid in order to be able to post these ads and that's where my question goes, is that because you're not viewed as a neutrality communication vehicle when someone posts an ad, your algorithms, targeting, insertion of your manner in which you access those that are accessing your content, also influences the process. what did you do when you received these payments for these ads? i was just in montenegro last year, where they were having a vote about nato.
2:58 am
they pointed to the billboard and said, that's a russian ad against nato. everyone knew it was a russian ad against nato. it was blindly placed on a billboard. you go further than just putting it on a highway. you look also to users and their individual interests in order to make sure they see targeted content. tell us, once you got the payments for these specific ads, we'll start with you concerning twitter, what was the activities twitter does, with the monies it received with respect to these russians ads in their attempts to influence. you didn't just post it on a highway, what did your company do with respect to the content of this ad and its direction. and i'd like you to tell us if you did, and what you did with respect to farming this out to your members. >> the russian today ads for example, when they were paid for and we were approaching them as
2:59 am
a news organization, they had a lot of options on our platform. they largely used promoted tweet. they take a tweet of a news story and promote it, so it is seen by users who don't follow them and potentially want to drive viewership to their own platform or have them followed back. they can target those ads based on geography. we saw very sergeanted ads u.s. citizens that focus other media sources. they were in florida and they do have those targeting capabilities. >> you're assisting them in that process, because they're coming to you as a customer. you assist them to identify where that's available to go? >> we show them tools available. here's how to use the dashboard. here's how to put out your tweets. we educate them to understand
3:00 am
the platform. >> you had the documentation as to how they were directed in this, on these particular ads that you were paid for that we have a concern about? stretch. >> congressman, all of the ads that were purchased were self-serve platform so there was no human interaction with any of the advertisers. the ads that were served, like all ads go through manual and automated review. it's in looking at those that were identified improvement in terms of tightening our ad content guidelines. >> my understanding according to the numbers we have it's somewhere around $100,000. that doesn't buy and ad across all of facebook. with your process -- i didn't say it was with an individual,
3:01 am
through your process there is a targeting selection that occurs, everybody didn't see these ads, how did facebook, once receiving these payments, take this content and determine who was and who was not to see it? >> the self-serve ad permit platform allows an advertiser to target it. a small subset of them were targeted to individual states, and then they were targeted at interests. once those targeted decisions were made, facebook, we look at our users and try to serve the information to users who have demonstrated the interests that have been selected by the advertiser. >> we did not have that information given to the committee and we have not heard yet from google. >> your time has expired. >> mr. carson? five minutes. >> thank you, sir.
3:02 am
thank you, chairman. >> russia made up groups like being patriotic and don't shoot cynically mimicked legitimate american organizations. we've established that, one fake group being patriotic, which amassed over 200,000 followers, pushed out images like these. it cynically ex-employeds grieving officers and their loved ones in order to pit americans concerned about our law enforcement personnel against americans concerned about african-american lives lost during police encounters. the second ad, fake group don't shoot describes police brutality. this masked over a quarter million followers on facebook made repeated ad buys displayed the page over
3:03 am
320,000 times across american facebook accounts. now, as a proud son of indiana and former police officer and member of congress, i reviewed these ads on facebook with a bit of disappointment, anger and concern. my concern is that a dictator like vladimir putin abused flaws in our social media platforms to inject the worst kind of identity politics into the voting decisions of at least 100 million americans and fear that we as a nation are not doing enough to identify continuing foreign digital interference on this important national debate. mr. stretch, did the russian ads like the ones exploiting violence between american citizens and law enforcement meet your criteria for rejection? >> congressman -- >> if not, why not?
3:04 am
>> so all of them had no place on our platform because they were first and foremost run by inauthentic actors and they should not have been on our site. they should not have been advertising on our site and we're investing to prevent this sort of behavior in the future. beyond that, many of them violated our policies that were in place at the time and should have been rejected through our customary ad content review. that has identified for us opportunities for improvement and investment in terms of making sure we have enough people and our systems are tuned enough to the guidelines. then, third, some of the ones that didn't violate our guidelines at the time, have caused us to tighten our guidelines because we saw some of these, we applied them to our guidelines, we thought, this is not stuff we want on facebook. we took a hard look at our policies, particularly around ads on divisive issues and
3:05 am
particularly around violence and we're turning the crank on those policies and will be applying those going forward. >> when an ad was by black matters opposed to black lives matters, did that trigger anything to your knowledge? >> not to my knowledge, congressman. >> how do you detect messages that foment violence? algorithm or internal controls? how does that work? >> with respect to advertising content it is a combination of manual and automated review. the cycle really we go through is we have a policy. we have people apply that policy to ads that come through for review and we train the system that only individuals can make on an unusual basis.
3:06 am
with respect to violence we're looking to brandishing a weapon as something that would be prohibited under our ad content guidelines, and then training our systems and using artificial intelligence to make sure that any ad that gets run through this system, that has a weapon being brandished gets at least surfaced for manual review, if not just banned outright. thank you, sir. mr. chairman, i yield back. >> the chairman yields back. dr. wenstrup. five minutes. >> i'm sure when you started your businesses as they grew it was to bring people together and not apart as the wright brothers never intended for the airplane to be a weapons of mass destruction. that's what we have to face in this world. it's more than just the united states we have a problem with russia media meddling, internationally, people that
3:07 am
have written things against the soviet union in russia have been attacked on social media and their lives destroyed in these processes, not just here in the states. i want to bring up something in the same vein but not political realm, it's been brought to my attention of innocent people being used to create a false persona and scams on dating sites and back themselves up by check me out on facebook. these are fake. sometimes they use the person's name and sometimes it's just their image. they're pretending to be someone else. i know you're working hard on these types of things. but that's not the reputation you want your businesses to be. i'm particularly talking about facebook. so if you could, maybe bring me up to speed on what you are
3:08 am
doing on that front as well. because i'm afraid what was probably intended to be for good use has been turned into a tool for nefarious behavior, in many fronts, not just political. >> thank you for the question, congressman. when i said earlier today we have a responsibility to address something like foreign interference on the platform, we view that responsibility broadly to prevent our platform from being abused, used for abuse of any kind. and certainly the sorts of safety considerations that you've identified are paramount. so any behavior that is intended to put people at risk is a concern of ours. and we have teams addressing it. on child safety, in particular, we have robust teams that investigate reports of child safety certainly, and that are also looking at behavior on the platform that is consistent with troubling behavior to warrant investigation. and whenever we see anything that looks like it may be leading to real world harm, we
3:09 am
reach out to law enforcement and make sure that people's physical safety is secured. and last thing i'll mention is here again as i alluded to earlier, we have a good track record of sharing information among the industry. no one in this industry wants to see their platforms used to put somebody in danger. and we have a successful record of sharing information, sharing threat information, and working with law enforcement effectively. and that is a good construct, i think, for how we think about the foreign interference threat going forward. >> with law enforcement, is that a two way street? are they coming to you saying look out for this type of thing or this type of image or this name or whatever the case may be? is this a two-way street that's getting more robust, i would imagine, or is it one way, you are contacting them? how does that work? >> in many instances, it's been a two-way street.
3:10 am
so take, for example, our work trying to keep terrorist work off the platform. we are able to provide information and expertise what we are seeing. and the government has done an effective job of letting us know what they are seeing and giving us threat signals. i think with respect to nation state actors, we've had historically an effective dialogue or threat sharing information with respect to traditional cybersecurity actors. and we are hopeful going forward with respect to this disinformation activity we'll establish the same sort of dialogue. >> it seems to me if someone is recreating a facebook page out of asia, and you look on the facebook page and they are creating a person that says they live in montana, there should be a red flag. and i'm just curious if you have ways of catching that automatically or do you have to comb through each one? >> most of the accounts, most fake accounts are caught automatically.
3:11 am
many millions of them. our systems catch most of them automatically. they generally do come from particular regions of the world. and they are generally financially motivated. what makes this -- and so those systems have been in development for sometime and they are effective. we continually have to improve them. what makes this threat so insidious is these were carefully constructed profiles that i this i were maintained and cure rated to appear authentic, including disguising in most cases, not all, geographic origin. it is very useful for us, as we think about how to detect this going forward, but we do believe we need to up our game. and we are working on it. >> mr. quigley, five minutes.
3:12 am
>> let's look at unpaid content for a second. sometimes these fake accounts are pulled down, but the fake story takes the false claims of widespread voter fraud, for example, generated by these accounts have spread thousands of thousands of times often picked up by legitimate news accounts. what do you do to flag that? what do you sense is your responsibility? before any of you answer, let me just note this, that if we are asking are we still in this situation. just a short time ago, and i'm talking about when this meeting started, on twitter if you clicked on the hashtag nyc terrorist attack which is quote trending marked with a red button saying, quote, live, the top tweet links to an info war story with the headline, imam i warned de blasio about new york city terror. he was too busy bashing trump.
3:13 am
this is a realtime example of when we talk about this information being weaponized. how quickly can you act? and what's your responsibility to set the record straight so that the people who saw this know that it's fake news? and at least at some point in time it can't keep spreading like a virus through the legitimate world. >> that's something we are thinking about all the time. because it's a bad user experience, and we don't want to be known as platform for that. and your example for example the sim krefl the system self-corrected. that shouldn't be the first tweet. it should be a u.s. article last time i checked. >> but you saw this? >> usa today. >> at lunch today i did. and i also saw the system correct it. >> can you give me a really good guess on how long it was top? >> we can follow up with you and your staff on that. and i don't have the stat in
3:14 am
front of me so i don't know. but like we said earlier we are trying to balance free speech with making the information you see on the system especially around trends that we direct you to, so if you are clicking on a hashtag, we want to make sure you are seeing verified accounts and accurate information and reporting. sometimes it doesn't work as we intend. we learn from those mistakes and tweak and modulate going forward. >> beyond the correction, do you have a responsibility to flag something as this was fake news? >> we see our users do that a lot. we are an open public platform with respected journalists and other organizations who point these things out. you may have seen that in this instance. >> someone is breaking the law, you must feel like you have a responsibility to do something about that? as you said, this is -- with this extraordinary gift, this platform of free expression comes the responsibility, you
3:15 am
all talked about. so if you know something is illegal you know you have the responsibility to do something. at what point does this become something where you can't just correct it? you've got to say to the public this isn't true? >> right. and we take swift action on illegal content and illegal activity on the platform. a good example of this is the text to vote. voter suppression tweets that we've turned over to this committee. we saw swift action of the twitter community on disputing those claims. and twitter actively tweeted once it discovered these things were on the platform to notify our users this was fake information. that you could not in fact vote by tweet. and pointing people to a tool that would allow them to find their nearest polling place. >> is that because it was illegal activity or if something is just fake, do you think you have an equal responsibility? >> we took that down because it was illegal voter suppression. we were actively working on how do we balance what is real and
3:16 am
fake. and what do we do in the aftermath of something being tweeted and retweeted like you said. and have people having seen it. and how do we make sure they are seeing other view points and other facts and other news stories. >> do you have a policy right now where if you know something is out there that's not true of saying so? >> we do not. we have a policy that fosters the debate on the platform. we have a policy that takes down a lot of that content because it comes from automated malicious accounts or spammers. that stuff we are removing and acting on as quickly as we can. >> and i understand how you are trying to distinguish that. but the fact is if something is fake it doesn't matter if it's from some bot or something, it's not true and wildly obvious before it goes viral and picked up legitimate, you must feel like you have some responsibility? >> we are deeply concerned about that. and figuring out ways we can do it with the right balance.
3:17 am
>> thank you. >> time expired. mr. stewart five minutes. >> thank you mr. chairman. i'll move very quickly. thank you for being here. i want to come back to the thing about fake news and it makes my head explode and i'm sure yours as well. if it doesn't, it should. i want to go through something you said. you said all the right things or said most of the right things. i want to try to put some numbers, some actual something we can measure to that. and go through two questions. they are almost yes-no questions. on a scale of 1-10, how confident are you we understand the problem? how confident are you on a scale of 1-10 that you understand how pervasive the manipulation of your platform has been by foreign agents? >> i'm confident that we found what we found to date. but there may be more so we'll keep looking. so an 8. >> an 8. okay. >> i would echo that degree of confidence in the sense we are continuing to investigate, including sharing threat information among the companies.
3:18 am
>> so an 8? >> on terms of the scale. in terms of the importance, i think we are at a 10. >> okay. mr. walker? >> that's fair. i agree. >> i hope you are right. i got to tell you, i think -- i don't know that i share your 8 is very high. partly sitting on this committee, i don't view anything of much degree of certainty any longer because we are continually surprised. but i do hope you are right. so saying we are at an eight or something close to that, what kind of resources have given not just to evaluating what's happened in the past, but to preclude it from happening in the future? and when i say "resources," i mean have you assigned a couple summer interns to fix this? or is this something you assigned a team of 20? a team of 100? what kind of resources have you given to fix this in the next, what i'm worried about is the next election. >> right. so coming out of the 2016 election and events around 2016 and misinformation and automated
3:19 am
account use, we declared safety abuse and information quality the first priority at twitter. and our ceo asked our engineers, our designers and product teams to drop everything they were doing and try to solve this problem. we believe we made meaningful improvements around looking at things like behavior and stopping malicious automated accounts. we have a team called information quality team dedicated to stopping this. >> i'm going to stop you. so you have dozen of people? >> we had at a time thousands. we have hundreds. and we are continuing to try to figure out what our resources are. >> significant resources then. facebook? >> congressman, today approximately 10,000 people at facebook work on safety and security across our products and security community operations team. >> on this problem. >> on safety and security generally. by the end of 2018, 20,000. so more than doubling those teams. >> okay. significant resources. mr. walker again? >> on this investigation i would say hundreds. on the broader question of
3:20 am
safety security, thousands. i think the broader question here is improved artificial intelligence machine learning algorithm to deal with this at scale. >> brings me to my last point, and this is so difficult so i'm glad i'm not sitting in your seat. all of you have used the term "fake news." it's been used pervasively in this hearing. and we all recognize fake news is in the eye of the beholder many times. there are some things that are reported that are demonstrable untrue. but the vast majority of it is there some spectrum of reality and opinion. and i'll use mr. quigley example, he said, among warren -- imam warns de blasio and he ignored it because he was too busy criticizing trump. there is element of it. perhaps imam did we don't know yet. and him advising trump to some degree is a matter of opinion. now to my friend mr. quigley,
3:21 am
and i don't mean this as a criticism, but to him that's fake news. someone else reads that and they see legitimate critique in there. how in the world do you intend to identify fake news without weaponizing this in the political realm? because, as i said, there is an enormous degree of opinion included in almost every bit of that. and if you are viewed as being political in this, and it's my fear you will be regardless of what you do, if you are viewed as being political, it's not monitoring fake news and it's editorializing it, and the best example of that is fact checking. fact checking is opinionated many times as anything else that we see. and i have 24 seconds. you can't answer that. so i'll express my fear of it. but if you have a quick response i would be interested in what that might be. i don't blame you for remaining silent. you understand though why we are concerned about that and the challenge. and we want to help you on that, but we live in a political world already. i hope we make it less political
3:22 am
than more so. i yield back, mr. chairman. >> jim yemced back. jim yields back. mr. swalwell, five minutes. >> thank you, mr. chairman. >> can each of you assure the american people that you have fully searched your platforms and disclosed to this committee every russian effort to influence the 2016 election? mr. edgett. >> we have provided everything we have to date and continuing to look at this. so there will be more information that we share. >> mr. stretch? >> the same is true. particularly in connection with, as i mentioned earlier, some of of the threat sharing that the companies are now engaged in. >> mr. walker? >> yes, we have done both the review of the leads we have received and generated ourselves and then cross checked against other indicators we developed. >> so there could be more to come? >> the investigations will continue. that's correct. >> over my right shoulder are two of the ads that have been disclosed. one was an rt 1 in evacueeing wikileaks. the other was an rt ad invoking
3:23 am
donald trump and the debate. can you give me a yes or no. have you run analysis as to whether these ads were posted in duplicate form from non-russia sources? and what i mean is knowing that the russians used cutouts sometimes, have you done a pixel analysis to determine whether some other source posted this exact same ad? yes or no. mr. edgett. >> i'm sorry the answer is not sure but i'll follow up with you. >> thank you. mr. stretch? and i'm talking about every ad that you've disclosed. have you done a duplicate analysis? >> i would have to follow up with you, congressman. >> thank you. mr. walker? >> same answer. >> and do you share a concern perhaps that analysis has not been done and the russians did use cutouts and there are far more ads out there that they used beyond the 120 plus million views that occurred than what we know about? >> we did try to link accounts and look across a number of identifiers. so it may have picked up things like this. we were trying to be as
quote
3:24 am
exhaustive in our search as we could. i can't answer the question if we looked to see if they were promoting the same news from other more legitimate looking sites. >> we'll follow up. thank you. do each of you believe your company's and other social media platforms have a duty going forward to report to the fbi if you see foreign election activity before they do? >> we are working with the fbi constantly and notifying them of the illegal activity or take threat information from them. >> mr. stretch, do you believe that duty now exists that we now know to warn them? >> yes, we think that's an area of potential improvement and hopeful we'll receive threat information that the bureau is aware of as well. >> thank you. mr. walker? >> again, i agree. >> and did any of your companies return ad revenue to the russians? meaning that did they make money on this? >> mr. walker?
3:25 am
>> so we in some cases we had sites such as rt that would show ads against their content and they made money from those ads. that's correct. the same is true beyond the internet because rt is on saturday light stations, cable, hotel television networks. they buy advertisements in magazines, airports, et cetera. >> so if i understand this right russia ran an interference campaign, attacked our democracy, sought to undermine our ability to choose, and they made money on it? >> so rt shows advertising on all the platforms we discussed. cable. satellite. internet. et cetera. and presumably, yes, they monetized those ads. >> mr. stretch? and mr. walker i mean did google pay rt part of your ad revenue back? >> google gets a small amount of it but some of it goes back to
3:26 am
the publisher. >> mr. stretch? >> the ads we were discussing were all paid for and didn't generate revenue on facebook. i would say approximately $100,000 and then quite a bit more we have contributed to the defending digital democracy project that's focused on in a bipartisan way on election security and protection. >> thank you. mr. edger? >> no, we didn't pay rt for any of this contempt. in fact, we have banned them as advertiser and donating the revenue we received to further education and academic research around the use of twitter. >> and mr. chairman, i'm entering into the records exhibits a and b additional facebook and instagram advertisements available to the public beyond what we show today and twitter handles turned over to use by the company connected to the research agency and ask for unanimous support to do that. >> and thank you again. i believe that our democracy was tacked by the russians in this last election. they infected our political process with a virus we have not yet kicked.
3:27 am
i believe the best anecdote is strengthen our defenses especially in the social media platforms that they used and we need your help to do that. >> the gentleman yields back. mr. crawford five minutes. >> thank you. i want to stretch with you and switch gears a little bit. a few weeks reports surfaced in multiple news outlets about facebook having desire to hire up to 1,000 additional personnel that would have top secret security clearance along with law enforcement agencies in the wake of russia's interference in our last election cycle. when i heard about that obviously i'm concerned because we have north of 17 agencies in the federal government right now that gather intelligence and analyze that information and those agencies already work with companies like yours and others here today. and i was also concerned by this report because as a committee, number one, we exercise over sight over those agencies to ensure they are doing their jobs and number two, that they're doing it legally.
3:28 am
so my question is the reach and impact that facebook has in the lives of every day citizens, the depth and breadth your platform has, average users somewhere in 50 hour week range. why do you think facebook has a need for 1,000 individuals with security clearances when we already have intelligence agencies doing that work at the federal level? can you answer that? >> congressman, we are not hiring 1,000 people with security clearances but we do require people with security clearances for two reasons. one, there is expertise that often comes with the security clearance that helps us understand threats like we are talking about today. second, and more concretely, there are national security related law enforcement process issues that we have to navigate. and we need personnel with security clearances to allow us to engage in the appropriate dialogue with the authorities. >> so 1,000, that's overstated.
3:29 am
how many would you estimate that you would be bringing on with security clearances? >> i don't have that number offhand. it would be in the single digits, potentially in the teens. >> hmm, that's interesting. that's not the report that i read. but that's okay. i'll take your word for it. what measures would you take to ensure that the american people can trust that you being social media platform can adequately do what you just described? and our colleagues on the intelligence committee and oversee that work can trust what you say and what you just described? >> i'm sorry, congressman, can you repeat the question. >> basically what i'm asking is you have these individuals, you are saying in the teens, not in the 1,000 range, as i read in a report, but in the teens, that you are acting in sort of interacting, if you will, with our existing intelligence community. what steps are you taking and
3:30 am
what does that interaction look like? and how can we be certain that in fact it's being done and we are able to exercise over sight over you if necessary? >> i understand. thank you for the question. the primary function we are describing involves the issuance of law enforcement process pursuant to statutory authority that puts safeguards in place for the potential subjects. for example, surveillance. i would think that through this committee's oversight of the intelligence community and exercise of those authorities would come with an understanding of how they engage with companies and how the companies themselves are responding to process. >> i really would probably like to get mr. walker, if you would weigh in on that same issue. i don't know if you have plans of hiring additional personnel with security clearances to the same end that we just discussed
3:31 am
with mr. stretch. but your thoughts on that? and that role going forward? >> sure. we have a very limited number of people with security clearances, again, to facilitate the kind of exchange with government as well as government contracting purposes. we think that exchange of information is quite valuable in terms of getting additional government leads to allow us to do our investigation as appropriate. >> mr. edgett, would you like to comment? >> we do similar sharing and have good working relationship with law enforcement on the ground on these issues. i don't know how many or if any have security clearances but we are also sharing information back and forth. >> you can probably imagine, you know, anticipate that folks as they understand that it's public that you have individuals within your platforms working for you that have security clearances and interacting with the intelligence community, that might give them pause.
3:32 am
do you see that folks average user might be concerned about that that you might then be perceived as a de facto intelligence agency? mr. stretch? >> i can understand the concern and i can assure you congressman, that any information we provide to the intelligence community as pursuant to lawful process that we examine closely. >> out of time. yield back. >> thank you chairman and thank you for your testimony today. i think all of you would agree that this activity is a grave threat to american democracy and our democratic processes. as three of the largest technology companies, you now have a responsibility to pool your expertise as the intelligence community did to produce vital assessment to initiate a joint investigation to uncover the full extent of russia covert activity on your platforms of the each company will need to
3:33 am
commit to an agreement to share information across companies. and we in congress can push the law enforcement to share information within their purview as well. like with counter terrorism efforts a two-way stream will be vital so that the companies can benefit from leads to inform their forensic informations and and future defensive efforts. and for the companies to alert the intelligence community and law enforcement to state sponsored foreign interference efforts so our agencies can have a robust comprehensive understanding of foreign intervention. will your companies breaking down any cooperation and personnel to a joint investigative initiative and producing a public report? sir? >> yes. >> certainly, congressman, we are investing heavily now and working with one another. and we think the public report will be the product of this committee's important work.
3:34 am
>> and i want to echo that. we have been exchanging leads with the other companies here and other companies as well. we welcome additional leads from government or other sources. >> thank you, gentlemen. are you also intending to turn over to the committee any kind of direct messaging that went on among the different accounts that were subject to this activity? in addition being able to buy ads for example, these accounts can send messages to other folks. or to each other. are you willing to turn over those direct messages? >> direct messages, i'm -- directing that at me since it's our product, are the private communication between our users so we take that privacy right and responsibility very seriously. so with the right legal process, we will work with law enforcement or others to provide whatever is necessary for an investigation. >> but do you see that as a legal issue within the united states? certainly you are not making the
3:35 am
argument that a russian account a fakely created account has some protection of privacy here? >> we are providing and using this rule across the board. we just require the right legal process to turn over information. some users may end up being fake, others will be real. so we take a principled approach and under the privacy laws here and around the world of making sure we are responding to the right legal process to turn over that kind of nonpublic information. >> twitter? facebook? >> congressman, we believe today we've responded to all of the committee's request for information and we've committed to full cooperation with the committee. the question private messages does indicate separate and perhaps thorny issues. if the committee does have a request for that sort of information we are happy to take a look and do what we can. >> and i know many of us would
3:36 am
like to see that ad and it could be vital in understanding exactly how this was carried out. and messages platforms on google? >> i join in that answer. in many cases the accounts we saw were being used actually to create social network accounts themselves. >> and then on facebook let me ask you, do you know whether any data lists were imported into facebook to do the targeted advertising from these accounts? >> the advertising was primarily as i mentioned earlier, fairly rudimentary. it didn't involve audience building in the way that many political campaigns have audience building. >> then let me just ask. did you check to seal whether there was any importation of data for targeting? the reason i ask is voter registration rolls in different states across the united states were hacked into. and we can't say for certain, but it's possible that somebody
3:37 am
stole information. >> with the ads that we provided to the committee, we provided all the targeting information. and we didn't see anything like that in that information. >> all right. thank you, chairman. i yield back. >> five minutes. >> thank you, mr. chairman. this conversation we are having today is october. we should have had this conversation a year ago. my questions will focus on the vetting process and the timeline. but broadly i want to start out and ask for each of your platforms, starting with twitter, how many total accounts are bots or trolls? not specifically what we are referring to today regarding ma lined russian influence but just generally how many million of accounts are bots or trolls? >> and we do regular audits and tests for that and determine that less than 5 percent for years have been false or spam accounts. >> so less than 5%. >> of all users.
3:38 am
>> what number is that? how many millions of users is that? >> it's 5% of about 330 million. >> okay. so facebook how many of your accounts, how many of your facebook pages are inauthentic facebook pages? >> so we measure this question by accounts. and we disclose with our financial statements each quarter our current assessment, will be providing an update assessment in a day or so when we file our 10-q. it's a small percentage. in the neighborhood of 2%. >> and that would be how many million accounts? or accounts, yes. >> we have over 2 billion users. >> significant? >> a lot. >> mr. walker, from google, how many inauthentic accounts? i know there were two referenced that were handed over to the committee but broadly, how many inauthentic accounts?
3:39 am
>> the two referenced today had to do with advertising accounts. so we have a continuing issue with trying to detect and deter fraudulent advertising on the system. as you recognize we are not a social network so our profile is somewhat different but it's a relatively small number. i don't have it in front of me but happy to follow up. >> thank you. my next question is the vetting process. mr. stretch, you've talked about the ad content review process. and i'm a fairly nimble facebook, twitter and google user. i think i'm the first person who signed up for facebook probably on this committee when i was in college. i know typically when you open up a facebook it's fairly easy and also easy to run ads. is the automatic response to put up the ad and review it after the fact? walk me through the specific ad content review. how many eyes are on that? what responsibilities do the ad content reviewers have? >> the ad will be reviewed before it's run.
3:40 am
and it will, based on its content, perhaps based on its targeting, it will either go fully through automated review. if the content and the targeting is something that we believe our systems are adequate to address on their own. so i'll give you an example. nudity is something that's relatively easy. it's prohibited and it's relatively easy for our systems to identify and prevent from running. there are other policies that are more nuance. so, for example, the difference between an ad that might have a weapon and an ad that might have a weapon being brandished. and that's a meaningful difference in our policies. so that one would undergo manual review. but in either case review occurs prior to the ad running. >> and the decision to not allow the ad to run or to take the ad
3:41 am
down after the fact, how quickly is that decision made? that's probably a different answer for each of those two groups. before it runs and after the fact? >> so our reviewers have the ability to action material when they are looking at it. really the question would be if the ad is already running, and it gets reported, for example, from our user community. if we miss something on the front end, oftentimes users will report it as violating. that drops it into a queue that then gets reviewed and again the reviewer can action the ad based on any policy violations. >> the ads turned over to the committee, this is a question for mr. edgett from twitter and facebook. what was the average amount of time when the account was opened to when the decision was made to close down the account or shut down the ads, what was that average amount of time, how many months? >> the accounts ran from june
3:42 am
2015 to august 2017. so there was a length of time for the accounts in their entirety. i don't have an average. we would be happy to do that analysis and come back to you. >> because that's an important question. election day is a specific date. polls shut down at a specific time. we live in a break neck media environment today. being able to identify and shut down those ads is important to solving this going into 2018. thank you. >> time expired. >> mr. heck, five minutes. >> thank you, mr. chairman. today's testimony is fully revealed how russian did a covert campaign to influence americans and distort our public debate and exploit elections by using soemced platforms every
3:43 am
single day. social media platforms every single day. this is no small feat and without effect. we have seen the effect they would go to do fake persona and news amplified by paid advertisements that reach more than 100 million americans in the run up to our presidential elections. this is no small feat. and it is not without effect. examples my colleagues presented today illustrate how the russians skillfully exacerbated some deep divisions that frankly haunt our country. stillfully. they put american against american and pushed extreme views on race and immigration and religion. and did this by pounding away again and and again on fears. this challenged the very notion of america. the most of us cherish. america that welcomes and treats all equally with respect. no matter your race or origin anyone or creed.
3:44 am
i ask you, raise your hand if you believe that all of this activity by the russians was without effect? me too. it was. with effect. so even as we continue to unravel and understand the full extent of the russian government covert mismisuse of why r platforms we don't have the luxury to focus on the past. january reads moscow will apply lessons learned from all campaign aimed at the u.s. presidential election to future influence efforts in the u.s. and worldwide including against u.s. allies in their election process. in fact, the ica identifies post election russian spear fishing campaign targeting u.s. government employees and think tanks. no in short russian operatives never left us. they're still in the house. in our house. the evidence you have unearthed
3:45 am
bears this out. i would like to leave you with three exhibits posted after the election. by the fake russian page stop all invaders. which boasts hundreds of thousands followers. shared hundreds of thousands of times collectively. to inject deeply bigoted antimuslim views into lt more than bloodstream. as debate raged on the ban. on december 19, stop ia posted this image. all face covering should be banned in all states across america and must not sacrifice national security to satisfy the demands of minorties. over 20,000 facebook users engaged with it. one way or another. on january 18 just days before the inauguration. this image posted with the text it should be obvious to every sane man, no place in civilized
3:46 am
society. more than 235,000 facebook users shared this post. and finally on february 20, the same post page post this picture of a handmade poster which generated 13,000 likes. and the text kick sharis out of america. the companies you have conducted of russia's exploitation of your platform is a necessary public service. it confirms the assessment and sheds any light on how expensive and con consistent the stealth russian operation has been. only scratched lt surface russian operatives worked across many online platforms repurposing content and messages and videos in order to extend their reach. burro more into american screens and news feeds to burr low into the psych. or to create an and inflame
3:47 am
prejudice. i too implore you to begin working jointly and discovering the full depth of how the russians weaponized your platforms. we need a total airing of everything that happened. and it needs to be comprehensive and thorough. and it can only be done if your companies work together and commit to working together. because frankly the stakes are high. the very health of our democracy is at rick here. if you this, if you commit to work together, if you work together, it will be no small feat. and it will be with effect. thank you, mr. chairman. >> mr. herd five minutes. >> thank you for being here. i want to pick up on a comment. that the gentlemen from washington just talked about. lessons learned. i had the honor of serving nine and a half years as an under cover officer in the cia. part of the time i chased a
3:48 am
russian intelligence officer across the world. and one of the things i want to learn the lessons you took from our elections, you were able to use some of the lessons learned in france and germany. i learned an intelligence game move counter move, move counter move. what were some of the counter moves we saw from the russians? after you took down some of the accounts and did we see a change in the tactics, technique and behavior. let's start with mr. stretch. >> some of the moves we made in response to what we have learned about o 2016 are to focus our automated systems on political trolling behavior. we talked about the total number of fake accounts on facebook for example the vast majority of those accounts are financially
3:49 am
motivated, involved in spam, and localized in the world. the activity we're talking about today is much more cue rated. not at scale. stst carefully maintained. and we have had to use much more subtle signals, things like the currency used to provide run a particular ad. or evidence of shared infrastructure. across multiple accounts. we have incorporated those in the system and believe we're affect effect. it's too soon to tell what the adversaries will do in response. i'm confident they'll do something. >> another question. so if you can keep the response concise. >> we use those two elections as opportunities to test the kbroouchts we are making to the system. and saw some very positive improvements in our ability to take down the malicious accounts. like you said they're getting better. we're having to look at these
3:50 am
things. and elections were a good opportunity. >> are you getting the kind of targeting data that would be helpful from federal government? when you found ut the internet research agency was that provided by the u.s. government or had to learn on your own? have you learned enough in order to purview your efforts? >> the ira tips were from news organizations in 2015. and also a third party company we use to do deep web monitoring to give ut threat. >> so commercial companies not from the u.s. government, would you like more help from the u.s. government. >> we welcome the help. >> let the record reflect the two other gentlemen are shaking their heads. can you see the two kpibexhibit here? one of them is the fake account
3:51 am
blacktist. the second one is south united her tej gn heritage not hate the south will rise again. do you think, i know you're not an expert, do you think the same person liked either one of those? >> i think they were directed at different audiences. >> the south has over 170,000 likes. what did the russians use -- what was their follow up. is this how they reached 126 million people by serving 80,000 posts after building the audience on each one of these locations? >> that appears to be the strategy. so they set up pages that were intended to appeal to different segments of the populous, they ran ads to try to drive
3:52 am
subscriber ship. >> and trying to erode trust. >> without question. all of the content we have seen today suggests that. >> i would say that, this activity by the russians is going to go down in history as the greatest covert action campaign in the history of of mother russia. not because of who won. but it drove a wedge wheen white house and the american people. and the intelligence services. it eroded trust in the public institution. like td press, congress and some of the great american companies. this is an attack and we have to work together. all of you said that. you can't do this alone. the folks up here are here are you. thank you for showing what the russians are trying to do when it comes to disinformation in the united states of the america. >> miss spear, five minutes. >> thank you to the witnesses
3:53 am
here. as i have listened to you today, i am reminded that america, we have a problem. we basically have the brightest minds of our tech community here, and russia was able to weaponize your platforms to divide us to dupe us, and to discredit democracy. i was impressed by one thing he said. he shut down rt. this question is for you mr. walker. rt russia today on your platform has 2.2 million subscribers. fox news on your platform has 740,000 subscribers. cnn has 2.3 million subscribers. the intelligence community assessment that was made public in january spoke about rt and
3:54 am
said rt conducts stra te-- why have you not shut down rt on you tube? >> thank you. we have heard the concerns. we spoke about this previously. the we recognize there are many concerned about rt slabted p perspective. at the same time it goes beyond the issue to cable and satellite tv and beyond. we have carefully reviewed the compliance with the policies. we haven't found violation of our policy against hate speech and violence and the like. >> it's a propaganda machine. the intelligence community all 17 agencies says it's an arm of one of our adversaries. >> and we agree. >> i would like for you to take that back to executives and rethink continuing to have it on
3:55 am
your platform. >> we agree the transparency is important for sources of information. we are working on additional ways to provide that. a range of government organizations. >> well, would you consider putting on that site that the intelligence community and the united states believes it's an arm of our adversary russia so people know what they're viewing? >> we'll lock at all forms of transparency sfl you said we tried to take things down as quickly as we can. the kremlin linked at tennessee gop. that twitter handle was very active as you know. and it was in fact a kremlin creation. the executive director of the real tennessee republican party stated he notified twitter that
3:56 am
the account was a fraud in september of 2016. and again in march and august of 2017. and it was not taken down until august of 2017. so would you adwree with me you did not take that down as quickly as you could have? >> absolutely. and that example alone caused us to relook at our policy and procedure. and i can say today that had we been reached out on that account today by the real tennessee gop, we would have taken it down much faster. >> congresswoman waters was targeted on that particular twitter handle. and viciously. i might add. would you provide us with a complete catalog of the tweets that came from the attack that might have targeted other members of congress and other groups? >> we can work with your staff on getting that information. >> one of the things with e noted during the campaign was
3:57 am
that often time the ads from the trump campaign mimic ad from the russians. i'd like to put up now an candidate trump ad. that included television ads questioning hillary clintons health and made it a feature of tweets and remarks. as you can see here in a tweet he posted at the end of august 2016. rt hamered the same message about clintons health. this rt advertisement less than two weeks after trump's tweet promoted video content produced by rt pushing the same message about clintons health. i'd like to understand who was mimicking who? have your investigations looked at whether the trump campaign was sharing russian content? have they looked at whether russians were sharing trump campaign content?
3:58 am
>> quickly gentlemen. >> we have provided all relevant information to the committee. you have access to broader set of information than any single company will. >> i agree with that. >> same for twitter. >> five minutes. >> thank you, mr. chairman. thaur thank you for coming today. there's been a theme on both sides of the aisle i want to sense a bit. if we can get clarity. at various points this afternoon one or more of you used words like authentic, accurate. misinformation, and disinformation. but at the same time at least one of you said you're not an ar by tor of the truth.
3:59 am
how can you not be an ash tor of the truth. if that it true, what's disinformation? >> disinformation we think of as inaccurate information spread with malicious intent by a foreign actor. that would not be permitted on our platform. we don't need to decide whether the information itself is true or false. to try to rid our platform of that. we don't want foreign actors. masker raiding something they're not in order to speak on politically devicive issues in the country. >> if the actor is not awe tentic the content whether it's accurate or not is immaterial. >> that's correct. >> and why would that analysis only be appropriate for foreign actors? why would that not be appropriate if -- there's been
4:00 am
discussion of voter suppression and information suppression. i don't know how people benefit from false information. so if that's the analysis for foreign actors why would that not be the appropriate analysis peri period across your platform in. >> everyone that shows up to facebook is required to be their authentic self-and most fake account activity is local to the country. >> those are two different things. i can be myself and say today is thursday. what are you going to do with that? >> we believe that you would be permitted to sa i that. >> under what constitution -- do you think the constitution protects intentionally false statements? >> sir, we are trying to provide
4:01 am
a platform or awe then tis si. >> i'm with you. i sim you're a lawyer. i know your colleague is a former ausa. i assume all of you are lawyers or you wouldn't have the jobs you have. so, is it constitutionally protected to utter an intentionally false statement? >> so, it depends on the con tek. this is precedent on that. on facebook. >> it is or is not? >> most cases protected. however on facebook, our job is not to decide whether content is trou or false. we do recognize that false news is real challenge, the way in which we're addressing it is by trying to disrupt the financial incentive of those profiting from it. where most of it comes from. most of this fake news problem is coming from low quality web
4:02 am
sites that are trying to drive traffic on every side of every issue. and by disrupting the financial incentive we're able to limit the distribution. we're also trying to make sure that users do know when a story has been disputed by a neutral third party. and alerting users to the fact. >> i'll stop there. i'm smiling because on the last break a couple colleagues and i were wondering who those neutral fact checkers are. i really do appreciate your desire to want to have a neutral fact checker. can you let me know who the folks are? i would be grateful. because people in my line of work might take exception with the neutrality of the fact checkers. if i understand you correctly the authenticity of the speaker is very important. the accuracy of the content less so? >> that's how we approach it. exactly right. >> all right.
4:03 am
for the life of me i do not understand how a republic is served by intentionally false information. i get it, you don't want to be the ash tor of opinion. don't want you to be either. but today is not thursday. if i say it is i swear i don't understand how fellow citizens benefit from me telling them something false. and i'm saying it with the intent to deceive. for the life of me i don't get it. but i'm out of time. >> second round. so with that, five minutes. we'll do one more round for each. >> thank you mr. chairman. we have agreed to release the facebook ads with the geographic targeting data. do either of the other two companies twitter google have objection to our releasing the ads identified as coming out of the russia as well?
4:04 am
>> no. i'm sorry, no. we don't have objection. we think that's part of the education we'd like to see for users. one of the reasons we're streaming today's hearing on twitter. no objections from us. >> mr. walker? >> no objections we're happy to work wtd committee on that. >> the broader 3,000 is a different conversation. >> correct. i want to make sure the companies. we have agreed to lease the facebook ads. they have been to be scrubbed for personal information. i want to make sure the companies are cope stettic with releasing their ads. >> with the private images we ask like you did with the boards to blur the faces out. >> i would ask also the exhibits we have used today and referred to be made part of the record.
4:05 am
couple quick follow ups, one my colleague asked if you would work on a joint report together. i'd like to under score my support for that concept. you are uniquely positioned better than us to be able to identify the interaction between your platform and how advertising on one platform led to likes that may have been used to target people on twitter. we're not in a position really to to do that. would you be willing to combine forces and share with us the report on the sum total of the russian social media interaction between your platforms during the campaign? >> we're sharing threat information all the time. and we can work together on this issue to get bert together. >> wuld you commit to providing us with a report that sets out
4:06 am
the length and breath of the russian use of the platforms? >> one of the things we're working on is a formal lized threat and information sharing body that will address cyber security threats generally and the goal is to publish among other things information. so we're certainly once with e get that body set up we will put that on the agenda as something to discuss. >> we already sharing with the other companies. information about google g mail accounts that might have been used to set up accounts on other services and we're happy to join in the joint effort. >> i appreciate that. i don't hear it as a firm yes. unless you want to somehow set us at loose within your data bases. we're not in a position to do what i'm asking you to do. you are. so i ask you to take that back and provide us an answer. i want to get back to the
4:07 am
question i poesed which is that is, would you agree that the effect of your al gor -- because type of things that go viral. because of the way you prioritize keeping eyes on the platform rather than showing people true information. the the truth is not what rises to the top of the feed. it's not the criteria that's used. what's the social responsibility here as one question. and the related question is do you feel an obligation to those who were influenced by the russian media you can identify to give them notice they were subject of russian sponsored covert advertising and
4:08 am
propaganda. they maybe future targets of it. as they have been identified by it by clicking on pages and following pages in the same way credit agencies have a obligation to notify their customers when the identity has been compromised. do you feel an obligation to notify your users they have been the subject of russian propaganda? >> given the i'll ask each witness to provide a written response to that answer. five minutes ask a question. so respect by the folks. i'll ask them to take that for the record. i recognize myself for five minutes. mr. walker, at the end of the conversation you talk about what you plan to do for 2018. creating a data base or whatever. with respect to allowing searches and that -- i guess who put up the ad who paid for it. would that be realtime or on a certain each week we have -- was
4:09 am
available or after the election is over? when will that tool be available? >> we're working out the details now. the goal is not after an election is over. but periodic reports. it's hard to compile the information realtime. periodically and with enough notice it would be useful in the electoral process. >> okay. thank you. i yield back. >> i yield two minutes of my time. >> the general woman wanted to give you an opportunity to respond to the questions asked. >> quickly. i think there's a distinction between google search. whose goal is provide accurate relevant information. and so social net work concerns we have been focussed on here today. we think the heart is soul of the product is provide useful and ak ral information to users. we do notify have a difficulty many users are not logged it. it's hard to give to know
4:10 am
exactly who has seen what. we have information about a given user is g mail account being hacked. we do and have for years provided information about the hack. >> in terms of the question of the division, and discourse in the country. the data is mixed about where that's coming from. or what the cause is. what we find on facebook is that it enables a net work of loose ties that expose people to a relatively broad diversity of information. we believe our obligation is to service information that is awe tentic as we have discussed and that does present a range of views one of the things we're working. >> my time is very limited. would you acknowledge that the way your al gor it m functions
4:11 am
has the effect to deepen the divisions. >> we recognize the concern. the data is quite mixed. we do recognize the concern. our goal certainly is not to deepen division. our goal is brng people together. >> do you feel an obligation to notify users that have been exposed to this and maybe a further victim of it because ip addresses have been captured snd. >> we have tried to provide notification through public blogs and a hard questions blog on the web site that addresses this. and we're committed to working with the committee to publicize all of the content we have seen. the question of individual notice is much more challenging. and we'd be happy to talk to you further about that. >> i yield back. thank you. >> i would like to sort of dig deeper about solutions. and perhaps ways for that we in congress can legislate better
4:12 am
legislate and help you in and be a partner and making sure we have transparency and accountability. so it is disturbing to me that political ads that are on television and radio and in print must explicit si say who is sponsoring the adst. yet on the internet on your platforms they don't have to do that. would you be in agreement if we did a bill, i know senator has a bill that's over on the senate side that i fully support, would you be in agreement to having those kinds of disclosure on your platform as well? yes or no. >> we are in agreement with the general direction of that. and announced we're setting up a transparency center to do that. >> same for facebook. >> and for us.
4:13 am
>> good. i know that when there were when you -- did your investigation you found out there was a link back to russia you took down the page. mr. stretch. is there not an obligation also to notify so for example on the rally example that i showed before, there were thousands of folks that responded to that. do you not have an obligation to let those folks know that that was a hoax? or at least inform them who was behind that sponsored advertisement? that you know it's misleading and -- >> thank you, congresswoman. we have tried today note iffy people about the issue broadly. through information on our web site. through our white paper last april. through the hard questions blog. and in working with the committee we're open to all of this information being released publicly. it's a much more challenging
4:14 am
issue to identify and notify reliablely people who may have been exposed to this convent on an individual basis. >> you do know the followers of these pages, right? that's information you collect. i'm in the asking about the multiplier effect. i'm talking about direct disearnable followers. and people who liked those pages. >> i understand the concern that the challenges that much of the data is old and much of the data about followers that we have been able to provide is the result of estimates and modelling. doing that reliablely. >> thank you all for being here. i say this with respect.
4:15 am
senator burr was right. i like the fact that you're here i respect that. and i think you're being forthcoming. but respect us or not, like us or not, we are the elected leaders of the country. and leaders of the social media plot form should be here too. this is that important. but maybe next time. let me just focus on a couple question things. on google point of view. you reference that the notion would be to say that there will be an icon on the ad to find out where it came from. don't you think that's more difficult and less likely that someone will do that than just being obvious like on a mail. says paid for by citizens for. >> for example the variety of ways to do this. this is a thing we want to explore. on the landing pages for any ad. you can have different kind of
4:16 am
disclosure for display ad or video. search ads. >> the most obvious means the most obvious to this person watching will be the most effective. >> sure. it comes at a trade off in terms of the right to free speech and free expression for political advertisers. we're trying to harmonize interest. >> when i take out an ad on tv, what's the difference? do you think there's something different about the internet where there's greater free expression there? or perhaps less ob tv? >> i think all the platforms, newspaper or broadcaster or the internet are committed the notion of transparency. implemented in different ways. >> rules should be the same and disclosure should be the same. >> wi agree. internet advertising is a dynamic. ads are being created on the fly. trying to figure out how to implement a transparency.
4:17 am
>> it said paid for by. i don't care how dynamic the ad can be. and less effective. obviously the person is running the tv ad isn't thrilled they have to have. that's why i paid for this ad. it's expensive. do you understand or look alike audiences used and tracking pixel imbedded in third party web sites to your understanding? >> for the ads that we have disclosed to the committee? >> so far, the ones you know about so far. >> i'm not aware of the techniques having been used. i would have to kmek to confirm. >> for all of you, what's the loikly hood there's a lot of the ads still out there that you haven't found? >> we feel we have done an extensive are view for the period. that we chose around the election. we're working on it. >> if you had to bet something
4:18 am
that you own that you really like, what are the odds there are adds out there. >> there's some likelihood. i feel like we have done an extensive review. >> something you really like. >> i share mr. edgers concern. and in particular in light of the ability to share threat information. it's possible there will be more that we discover. >> that could be there now. out there now, right? those ads are on the facebook. >> certainly not associated with this cluster of accounts. we are focusing significant efforts on prevents this sort of bhair going forward. behavior going forward. perfection will be difficult. we'll be better. >> so we like wise feel we have done an extensive investigation. we don't know what we don't know. >> why did it take so long to
4:19 am
find them? this is a long time after. why did it take so long, oh look. >> we found a number of throughout the course of the last several years we found a large number of cyber attacks and the like. we addressed those as we found them. i was really only after the intelligence committee came out with the report. that we did a deep dive. it's true of the other companies. in this particular area. and found out more. >> is that the same true for facebook, that's why you didn't look so extensively, because intelligence community didn't come out with the report? >> it's fair to say our efforts with respect have been focussed on traditional cyber security threats. the intelligence assessment was an important piece of informs that caused us to look further. we published a white paper discussing certain activities in april and continued the investigation. >> thank you. >> five minutes.
4:20 am
>> thank you, chairman. thank you for the second round. do you believe russian intelligence services are still today on your platform? >> i think there's a possibility. we're working to find them. >> mr. stretch? >> the same is true. >> mr. walker? >> yes. same. >> twitter recently decided not to rin rt ads on its platform. is your assess. rt pushed fake news on twitter. >> that was part of the assessment. >> was part of the assessment they pushed stolen information from the dnc and the clinton e-mails? >> that wasn't part of the assessment. no. >> you agree that rt was pushing that propaganda? >> i saw them reporting on that and some promoted tweets. >> going back to the earlier conversation, from the disclosures we have from google. we have 1,100 videos that rt
4:21 am
posted. spanning 43 hours of content on you tube. and you mentioned earlier that some money was returned through ad revenue sharing to rt. how much money was rurp returned to rt as it related to ads that related to election interference? >> if i could clarify the 1,100 videos were separate. those were ads connected with this other disinformation campaign. or deceptive ads with no disclosure source. there was a minimal amount of advertising rove knew received. >> do you have a number? >> we can provide that information. it was minimum i believe. >> do you understand the concern that ad revenue that the russians are receiving back from u.s. companies as they interfere with our campaign could be used to buy ads on platforms like facebook and twitter.
4:22 am
and you essentially have a feed back loop to continue to try and interfere? >> yes. certainly there's a risk any time anybody can misuse the proceeds of advertising. >> having reviewed the propaganda ads that each of you disclosed, would you agree as it related to candidates trump and clinton, that the theme generally went like this. if they invoked either candidate the ads were protrump, and anticlinton. would you agree? >> that's generally what we saw. >> mr. stretch? >> for the ads that ran prior to election, that's generally accurate. >> mr. walker. >> we did not see -- we had a very limited of ads. i would describe socially devicive. rather than proor antieither
4:23 am
candidate. >> i want to point out we found propaganda that was expressed by the russians not just during the general election campaign. but also during the primary. here's a july 21st, 2016 ad. at this point donald trump is about to be the nominee for the republican party. there's questions about whether it will be contended. at the rnc convention. and rt is propagating a ted cruise story about ted cruise being booed for speaking as if he's running for president. and so i guess i would point out to my republican colleagues, that if this sper feerns campaign has taught us anything is that the russians don't care. they're not prorepublican. they're not antidemocrat. they're just pro-russian. and in this election they happen to find a candidate who is also very much pro-russian.
4:24 am
and so i hope that moves all to understand that in the next election, for whatever reasons the russian maybe be antirepublican. and that my republican colleagues too could be subject to attack. just as we saw in a smaller less significant way. that one of their own presidential candidates was part of russia's propaganda campaign. i yield back. >> mr. heck, five minutes. >> i have a simple question for each of you. do you acknowledge and recognize that the magnitude and nature of efforts under taken by the russians to interfere in the elections last year constitute an exten shl and material threat to the health of the democracy? >> we recognize that and
4:25 am
obviously a threat to all of our platforms. >> mr. stretch? >> yes. >> the activity that we saw in the platform was limited. we're not in position to take a broader view. when you look at the whole picture. >> you do not acknowledge is represents an exten shl. >> any effort to interfere with integrity is a significant material problem. absolutely. >> so, two things occur to me. with respect to that. and thank you. the first of which is that the actions that follow ought to be correspondingly significant to your acknowledgment that are very democracy has been threatened. and will be threatened going into the future. because it's the ic said it ain't over. and as i said earlier they're in
4:26 am
our house. i would use that as an opportunity to reiterate my call for ewe to work together to fully reveal and disclose the nature and at the point of this. and to come forth with continuing recommendations about how it is we can arm ourselves and prevent this in the future. and the second thing i want to point outs that i think it's important that you hear, is kind of the irony of this. you represent three of the best of the traditions of america. entrepreneurialism. you were all born here. not you individually. but companies. you flourished here. unless you want to be a part of clep tock si. dancing to the tune of the vladimir putin. which i'm invoking literally and figuratively. i hope that your actions coming out of the day match what you have acknowledged is the
4:27 am
seriousness of the threat. truly. our very way of life is at risk. just as truly you have a critical rule in safeguarding that. i hope you will. with that i yield back. >> miss spear. five minutes. >> thank you begun fagain for being here. can you tell us when you became aware that your platforms were being exploited by the russians? >> sure. it's very difficult to answer because it's many years we have seen a variety of cyber espionage. that really was we started a deeper dive at the intelligence committee report. >> so, after the election? >> that's when it became a focus.
4:28 am
yes. >> mr. stretch? >> like google, we have seen nation state actors including actors that we believe are connected today russia. trying to operate on the platform for sometime. and we have security measures to address that. in the summer the late summer of 2016 we identified an account that was this appeared to be involved in spreading disinformation. connected to some of the stolen account contents. we disabled that page shortly after -- that account shortly after discovering it. and subsequently communicated that information to law enforcement. following the assessment like google we under took a broader review of disinformation operations on the platform. >> thank you. >> similar to my colleagues wechb fighting these types of
4:29 am
issues for a while. we saw in 2015 ira activity and took large scale action against the accounts and shared the information with other companies at the time. coming out of the election we dnts learn about the use of the platform in the ways we're talking about now. until the intelligence committee report. and we're focussed on looking forward as elections continue to solving the problem going forward. >> which would suggest maybe the intelligence community should have been in contact with you when they first became aware of it. that's another question. mr. stretch, you have a lot of great play by the digital team of the trump campaign publicly. when they credited facebook with making a significant contribution to the trump victory. by pushing out data which included hundreds of thousands of variations of ads. micro targeted by geographic location and demographic. you had i believe a number of facebook employees that were imbedded within the trump campaign. how many did you have?
4:30 am
>> there was a team that was supporting the trump campaign the number varied over the course of the campaign. it was a happenedful led by one individual. who whom that was his primary assignment. i would ad this is consistent with the support that we offer any large advertiser including other political campaigns. >> in this case they were actually imbedded within the campaign. we have testimony from others to skugt that was true. >> with any of the large advertisers, our sales support teams will regularly be on site. and that was true in this case. >> so did facebook employees choose dh capability to employ to maximize the reach and exact of the trump campaign ads? the rule of the sales support team generally is to advise the client on the tools that are available to help them meet their objective. in the case of the trump campaign the objectives on
4:31 am
facebook involve fundraiser. the role of the team as it would be for the support of any campaign was to offer advise as to which tools help them meet objectives. >> so, again. you were successful. one of the concerns i have moving forward is how can we be assured that you will be using the power knowing that you and you alone understand your what creates optimal engagement. how you will encourage voter participation in a neutral fashion. have you given thought to that? >> thank you, congresswoman. it's an excellent question. one of the things we take a great deal of pride is in the role we have played to increase voter registration. and increase voter turn out. something we focus on in the u.s. on in every election as well as other elections around the world. so in the u.s. election for example, we believe our efforts on voter registration led to 2
4:32 am
million additional voters. registering across the united states. for the presidential election. we apply those techniques, those voter registration techniques as well as the election day encouragement to vote. on a neutral basis. >> but the extent to which face bok has hired to do ads. you would provide those services to one candidate vs. another? >> we have a comp comp plins team that trains representatives to comply with all federal election law guidance in the area. >> closing comments. mr. shift any comments you would like to make. >> thank you. thank you for coming in today. we appreciate your testimony. and know that a lot of the issues that we have been grappling with today are not
4:33 am
easy. and i think we understand we'll have to work to establish better lines of communication from the intelligence community to the tech sector to share insights they gain. so you can help fair it out foreign bad actors of using your platforms. we'll have to continue to have a dialogue in efforts to resolve them and provide appropriate over sight. we want you to be a successful. we need you to be successful. technology is one of the come pettive economic advantages we enjoy the rest of the world. in california we're particularly proud of the tech sector. i hope you take questions in the spirit in which they're intended. we'll continue to hold your feet to the fire. i look forward to further exploration of the issues. >> thank you. gentlemen i thank you for being here today.
4:34 am
i take it as my own personal responsibility as to who voit for, how i make the decision. what input i get and who i read. what influences are out there to make sure i understand who i'll vote for. against and for. and i would hope that every american would take that same position that is their own personal responsibility to not be misled. that being said, anything you can do and everything you can do to help me with that role of not being misled by folks who are tempting to do that is was appreciated and quite frankly unexpected. that you would move forward on this issue. the 2018 election is not that far out. the bad news is i suspect our adversaries will learn all of the things that you're doing to fix what happened behind us. and will look forward to trying to escape your fixes going forward. this is not one of the excuses that is one and done fix. it will be a constant moving
4:35 am
forward and moving battle. we have a role in making that happen. i would hope that my colleagues defense of the voter and not being misled and lied to would apply to making sure that when voters show at the poll they are a voter who is supposed to be there. and have an ability to determine the voter should be there and only vote once. the same kind of f. again thank you. you have a lot of work to do ahead of us. i appreciate the resources your companies and shareholders will put forward to make this get better. obviously adversaries will get better as well. with that, we are adjourned. thank you.
4:36 am
[ inaudible conversation ]
4:37 am
c-span washington journal. live every day with news and policy issues. that impact you. coming up tuesday morning, california democratic
4:38 am
congresswoman talks about her legislation to strengthen laws preventing and reporting sexual harassment on capitol hill. also political trade reporter discusses president trumps trade policy goals during his trip to asia. and live in baton rouge for the next stop on the c-span bus 50 capitols tour with louisiana secretary of state. and louisiana attorney general jeff land ri. talk about security of voting systems and public policy issues in the state. watch the washington journal live at 7 eastern tuesday morning. join the discussion. the house ways means committee continues work on the gop tax reform bill tuesday and wednesday. watch live coverage tuesday at 10:00 a.m. eastern. on c-span 3. and c-span.org or listen live on the free c-span radio app. up next, the head of the food and drug adminisat

156 Views

info Stream Only

Uploaded by TV Archive on