Wri en Testimony of Derek Slater  Director Information Policy Google LLC    House Commi ee on Homeland Security  “Examining Social Media Companies' E o s to Counter Online Terror Content  and Misinformation”  June 26 2019    Chairman Thompson Ranking Member Rogers and distinguished members of the  Commi ee Thank you for the oppo unity to appear before you today I appreciate  your leadership on the impo ant issues of radicalization and misinformation online   and welcome the oppo unity to discuss Google’s work in these areas     My name is Derek Slater and I am the Global Director of Information Policy at Google   In my role I lead a team that advises the company on public policy frameworks for  online content -- including hate speech terrorism and misinformation Prior to my role  at Google I worked on internet policy at the Electronic Frontier Foundation and at the  Berkman Center for Internet and Society     At Google we believe that the Internet has been a force for creativity learning and  access to information Suppo ing this free ow of ideas is core to our mission to  organize and make the world’s information universally accessible and useful We build  tools that empower users to access create and share information like never before —  giving them more choice oppo unity and exposure to a diversity of opinions   Products like YouTube for example have expanded economic oppo unity for small      businesses to market and sell their goods have given a ists creators and journalists a  pla orm to share their work connect with an audience and enrich civic discourse and  have enabled billions to bene t from a bigger broader understanding of the world     While the free ow of information and ideas has impo ant social cultural and  economic bene ts there have always been legitimate limits even where laws strongly  protect free expression This is true both online and o especially when it comes to  issues of terrorism hate speech and misinfomation We are deeply troubled by the  increase in hate and violence in the world pa icularly by the acts of terrorism and  violent extremism in New Zealand We take these issues seriously and want to be a pa   of the solution     This is why in addition to being guided by local law we have Community Guidelines  our users have to follow We also work closely with government industry and civil  society to address these challenges in pa nership within the United States and around  the world In my testimony today I will focus on two key areas where we are making  progress to help protect our users i on the enforcement of our policies around  terrorism and hate speech and ii in comba ing misinformation broadly     Enforcement on YouTube for Terrorism and Hate Speech    We have rigorous policies and programs to defend against the use of our pla orm to  spread hate or incite violence This includes terrorist recruitment violent extremism   incitement to violence glori cation of violence and videos that teach people how to  commit terrorist a acks We apply these policies to violent extremism of all kinds   whether inciting violence on the basis of race or religion or as pa of an organized  terrorist group     2    Tough policies have to be coupled with tough enforcement Over the past two years   we have invested heavily in machines and people to quickly identify and remove  content that violates our policies against incitement to violence and hate speech     1 YouTube’s enforcement system sta s from the point at which a user uploads a  video If it is somewhat similar to videos that already violate our policies it is  sent for humans to review If they determine that it violates our policies they  remove it and the system makes a “digital ngerprint” or hash of the video so it  can’t be uploaded again In the rst qua er of 2019 over 75% of the more than 8  million videos removed were rst agged by a machine the majority of which  were removed before a single view was received    2 Machine learning technology is what helps us nd this content and enforce our  policies at scale But hate and violent extremism are nuanced and constantly  evolving which is why we also rely on expe s to nd videos the algorithm might  be missing Some of these expe s sit at our intel desk which proactively looks  for new trends in content that might violate our policies We also allow expe   NGOs and governments to notify us of bad content in bulk through our Trusted  Flagger program We reserve the nal decision on whether to remove videos  they ag but we bene t immensely from their expe ise   3 Finally we go beyond enforcing our polices by creating programs to promote  counterspeech on our pla orms to present narratives and elevate the voices  that are most credible in speaking out against hate violence and terrorism    a For example our Creators for Change program suppo s creators who  are tackling tough issues including extremism and hate by building  empathy and acting as positive role models There have been 59 million  views of 2018 Creators for Change videos so far the creators involved  have over 60 million subscribers and more than 8 5 billion lifetime views  of their channels and through ‘Local chapters’ of Creators for Change   creators tackle challenges speci c to di erent markets    3    b Alphabet’s Jigsaw group an incubator to tackle some of the toughest  global security challenges has deployed the Redirect Method which  uses Adwords targeting tools and curated YouTube playlists to disrupt  online radicalization The method is open to anyone to use and we know  that NGOs have sponsored campaigns against a wide-spectrum of  ideologically-motivated terrorists      This broad and cross-sectional work has led to tangible results In Q1 2019 YouTube  manually reviewed over 1M suspected terrorist videos and found that only fewer than  10% 90K videos violated our terrorism policy Even though the amount of content we  remove for terrorism is low compared to the overall amount our users and algorithms  ag we invest in reviewing all of it out of an abundance of caution As comparison  point we typically remove between 7 and 9 million videos per qua er—a fraction of a  percent of YouTube's total views during this time period Most of these videos were  rst agged for review by our automated systems Over 90% of violent extremist  videos that were uploaded and removed in the past 6 months Q4 '18 Q1 '19 were  removed before receiving a single human ag and of those 88% had fewer than ten  views     Our e o s do not end there We are constantly taking input and reacting to new  situations For example YouTube recently fu her updated its Hate Speech policy The  updated policy speci cally prohibits videos alleging that a group is superior in order to  justify discrimination segregation or exclusion based on qualities like age gender   race caste religion sexual orientation or veteran status This would include for  example videos that promote or glorify Nazi ideology which is inherently  discriminatory It also prohibits content denying that well-documented violent events   like the Holocaust or the shooting at Sandy Hook Elementary took place We began  enforcing the updated policy the day it launched however it will take time for our  4    systems to fully ramp up and we’ll be gradually expanding coverage over the next  several months     Similarly the recent tragic events in Christchurch presented some unprecedented  challenges and we had to take some unprecedented steps to address the  unprecedented volume of new videos related to the events--tens of thousands   exponentially larger than we had ever seen before at times coming in as fast as one  per second In response we took more drastic measures such as automatically  rejecting new uploads of clips of the video without waiting for human review to check  if it was news content We are now reexamining our crisis protocols and we've been  giving a lot of thought to what additional steps we can take to fu her protect our  pla orms against misuse Google and YouTube also signed the Christchurch Call to  Action a series of commitments to quickly and responsibly address terrorist content  online The e o was spearheaded by New Zealand’s prime minister to ensure another  misuse of online pla orms like this cannot happen again     Finally we are deeply commi ed to working with government the tech industry and  expe s from civil society and academia to protect our services from being exploited  by bad actors During Google’s chairmanship of the Global Internet Forum to Counter  Terrorism over the last year and a half the Forum sought to expand its membership  and to reach out to a wide variety of stakeholders to ensure we are responsibly  addressing terrorist content online For example we hosted a summit in Sunnyvale so  G7 security ministers could hear the concerns of smaller pla orms We have also  convened workshops with activists and civil society organizations to nd ways to  suppo their online counter-extremism campaigns and sponsored workshops around  the world to share good practices with other tech companies and pla orms     Combating Misinformation    5    We have a natural long-term incentive to prevent anyone from inte ering with the  integrity of our products We also recognize that it is critically impo ant to combat  misinformation in the context of democratic elections when our users seek accurate   trusted information that will help them make critical decisions We have worked hard to  curb misinformation in our products Our e o s include designing be er ranking  algorithms implementing tougher policies against monetization of misrepresentative  content and deploying multiple teams that identify and take action against malicious  actors At the same time we have to be mindful that our pla orms re ect a broad  array of sources and information and there are impo ant free-speech considerations   There is no silver bullet but we will continue to work to get it right and we rely on a  diverse set of tools strategies and transparency e o s to achieve our goals      We make quality count in our ranking systems in order to deliver quality information   especially in contexts that are prone to rumors and the propagation of false  information such as breaking news events The ranking algorithms we develop to that  end are geared toward ensuring the usefulness of our services as measured by user  testing The systems are not designed to rank content based on its political  perspective      Since the early days of Google and YouTube some content creators have tried to  deceive our ranking systems in order to increase their visibility a set of practices we  view as a form of spam To prevent spam and other improper activity during elections   we have multiple internal teams that identify malicious actors wherever they originate   disable their accounts and share threat information with other companies and law  enforcement o cials We will continue to invest resources to address this issue and to  work with law enforcement Congress and other companies      In addition to tackling spam we invest in trust and safety e o s and automated tools  to tackle a broad set of malicious behaviors Our policies across Google Search   6    Google News YouTube and our adve ising products clearly outline behaviors that are  prohibited such as misrepresentation of one’s ownership or primary purpose on  Google News and our adve ising products or impersonation of other channels or  individuals on YouTube We make these rules of the road clear to users and content  creators while being mindful not to disclose so much information about our systems  and policies as to make it easier for malicious actors to circumvent our defenses      Finally we strive to provide users with easy access to context and a diverse set of  perspectives which are key to providing users with the information they need to form  their own views Our products and services expose users to numerous links or videos  from di erent sources in response to their searches which maximizes exposure to  diverse perspectives or viewpoints before deciding what to explore in depth In  addition we develop many tools and features to provide additional information to  users about their searches such as knowledge or information panels in Google Search  and YouTube      Conclusion    We want to do everything we can to ensure users are not exposed to content that  promotes or glori es acts of terrorism Similarly we also recognize that it is critically  impo ant to combat misinformation in the context of democratic elections when our  users seek accurate trusted information that will help them make critical decisions   E o s to undermine the free- ow of information is antithetical to our mission We  understand these are di cult issues of serious interest to the Commi ee We take  them seriously and want to be responsible actors who are a pa of the solution      We know that our users will value our services only so long as they continue to trust  them to work well and provide them with the most relevant and useful information We  7    believe we have developed a responsible approach to address the evolving and  complex issues that manifest on our pla orm     We look forward to continued collaboration with the Commi ee as it examines these  issues Thank you for your time I look forward to taking your questions   8