Jack Dorsey Testifies on Twitter's Response to Foreign Influence Operations

Twitter's Jack Dorsey testified at the September 5, 2018 hearing "Foreign Influence Operations’ Use of Social Media Platforms". Sheryl Sandberg of Facebook also testified. Google was invited but declined to attend. Here is a brief outline of Mr. Dorsey's opening statement:

I. Opening

• Twitter strives to be a global town square
• Must be a trusted and healthy place
• Abuse, malicious automation, and manipulation detract from it
• Never more important than during elections
• Share concern with committee about malicious foreign influence
• Hold themselves publicly accountable for progress

II. Russian influence in 2016 election, lessons learned

A. Retrospective review

• Twitter conducted a comprehensive review of 2016 election activity 
• Identified 50,258 accounts that were automated, linked to Russia, and tweeting election-related content, representing 0.016% of total accounts
• These accounts constituted 1.00% of election-related tweets; totaling 2.12 million tweets
• Twitter barred paid ads from two of the most active accounts generating paid content; affiliated with Russia Today (“RT”)
• Donating the $1.9 million that RT spent globally on advertising to academic research into election and civic engagement

B. Insights from review

• Although volume of malicious activity relatively small, any such activity on Twitter is unacceptable
• Recognize that as a private company, there are threats they cannot understand and address alone; must continue to work with elected officials, government partners, outside experts, etc.

III. Improvements

• Twitter recently developed more than 30 policy and product changes
• Made a number of improvements specifically for the 2018 election

A. Malicious activity detection and safeguards

• Refined detection systems; prioritizing suspicious account activity identification, such as exceptionally high-volume tweeting with the same hashtag or mentioning the same @handle without a reply from the account being addressed
• Increased use of challenges intended to catch automated accounts, such as reCAPTCHAs
• Implementing mandatory email or cell phone verification for all new accounts
• Removing 214% more accounts year-over-year for violating platform manipulation policies
• Thwarting 530,000 suspicious logins a day
• Received approximately 25,000 spam reports per day in March of this year; that number decreased to 17,000 in August
• Removed locked accounts from people’s follower counts, to ensure these figures are more reliable; accounts are locked when systems detect unusual activity and force a password change or other challenge

B. Analytic team

• Created an internal analytical team to monitor site and platform integrity
• Can respond immediately to escalations of inauthentic, malicious automated, or human-coordinated activity
• Also receives and responds to reports from across the company and external third parties
• Primary focus is election readiness

C. Political conversations dashboard

• Developed political conversations dashboard to evaluate the integrity of political conversations on the platform in the aggregate
• Dashboard surfaces information about sudden shifts in sentiment around a specific conversation, suggesting a potential coordinated campaign of activity, as well as information about groups of potentially linked accounts that are posting about the same topic

D. Candidate verification

• Verification, a blue badge that appears next to a person’s Twitter handle, lets people know that accounts of public interest are the authentic accounts (as opposed to impersonation or parody accounts)
• Making a concerted effort to verify all major party candidates for both federal and key state positions

E. Election labels

• Developed a new U.S. election label to identify political candidates
• Label includes information about the office the candidate is running for, the state the office is located in, and the district number, if applicable
• Features are designed to instill confidence that the content people are viewing is reliable and accurately reflects candidates’ and elected officials’ positions and opinions

F. Advertising

• Implemented updated Political Campaigning Policy; under the revised policy, advertisers who wish to target the United States with federal political campaigning ads are required to self-identify as such and certify that they are located within the U.S.
• Foreign nationals will not be permitted to serve political ads to individuals who identify as being located in the U.S.
• Twitter accounts that wish to target the U.S. with federal political campaigning advertisements must also comply with a strict set of requirements
• Accounts serving ads with content related to a federal political campaign will now be visually identified and contain a disclaimer
• In June, launched the Ads Transparency Center, which is open to everyone on Twitter and the general public, and currently focuses on electioneering communications
• Recently announced the next phase of efforts to provide transparency with the launch of a U.S.-specific Issue Ads Policy and certification process

G. Engaging key stakeholders 

• On August 24, 2018, Twitter hosted industry peers to discuss data sharing about hostile foreign actors and 2018 election security
• Have well-established relationships with law enforcement agencies active in this arena, including the Federal Bureau of Investigation Foreign Influence Task Force and the Department of Homeland Security’s Election Security Task Force
• Recently updated a Partner Support Portal; goal is to expedite response to reports from people active in the election arena. This includes election support organizations, U.S.-based research organizations, and universities and academics who study the spread of misinformation in the media. Reports from accounts within this select group are expedited and can be actioned promptly
• Partnered with experts at the University of Oxford and Leiden University to better evaluate work on conversation health

H. Accessing public tweet data 

• Recently updated developer policies, which govern access and use of public Tweet data made available to developers and other third parties through their application programming interfaces (“APIs”)
• Between April and June 2018 alone removed more than 143,000 applications determined to be in violation of developer policies; most violated policies against producing spam via APIs
• Continue to improve detection tools to stop misuse of public Twitter data
• Twitter now reviews and conducts compliance checks of all developers’ stated use of the data that they wish to access

IV. Recent activity 

A. Russian Internet Research Agency (IRA) activity 

• As reported to the Committee in January 2018, continue to identify accounts that they believe may be linked to the Internet Research Agency (“IRA”) 
• These accounts used false identifies purporting to be Americans, and created personas focused on divisive social and political issues
• The accounts represented both sides of the political spectrum
• Continue to work with law enforcement partners on this investigation

B. Iran activity 

• Twitter has been in close contact with our industry peers on this matter and received detailed information from them about the malicious accounts located with Iran, which has assisted us in our investigation, and we have shared our own details and work with partner companies