The X and Twitter information operations archive

What the archive was

X says it launched the first archive in the industry of potential foreign information operations in October 2018 ¹. Twitter wrote on 2 December 2021 that it had published the first public archive of data related to state-backed information operations in October 2018 ². By then it had shared 37 datasets of attributed platform manipulation campaigns originating from 17 countries ².

The X Moderation Research Consortium page gives the same total for October 2018 through early 2022. It says the datasets span more than 200 million posts and nine terabytes of media ¹.

What the data contains

The datasets cover persistent platform manipulation campaigns and information operations that violate X's platform manipulation and spam policy ¹. They include profile information, posts and media from accounts linked to state-backed information operations. Deleted posts are excluded ¹. X says the consortium data is not hashed, unlike the public archive ¹.

The policy behind the removals prohibits attempts to artificially influence conversations through multiple accounts or fake accounts. It covers inauthentic engagement, buying or selling engagement, and mass registration of accounts ⁴. For severe violations, X says accounts will be permanently suspended at first detection ⁴.

The move to a research consortium

On 2 December 2021 Twitter said it would launch the Twitter Moderation Research Consortium in early 2022. It described it as a global group of experts from academia, civil society, NGOs and journalism studying platform governance ². Global partners named in an August 2022 update to that post were the Stanford Internet Observatory, the Australian Strategic Policy Institute and Cazadores de Fake News ².

On 22 September 2022 Twitter opened membership to global researchers from academia, civil society, NGOs and journalism. It said the consortium held 52 datasets spanning nine terabytes of media and more than 220 million Tweets ³. It said it would keep sharing data on removed networks with their presumed country of origin, but would no longer share its own attribution information for these datasets ³.

Who can apply

• Applicants need a primary affiliation with an academic, journalistic, nonprofit or civil society research organisation ¹.
• Master's and PhD students qualify. Undergraduate students are not eligible at this time, according to X ¹.
• Applicants need data analysis skills, a specific public-interest research use, privacy and security safeguards, and a signed data use agreement ¹.
• People whose primary affiliation is industry or government, or who plan to share the data with governments or other outside parties, are not eligible ¹.

Reading the datasets

A dataset reflects Twitter's or X's own enforcement decision under its platform manipulation policy ^1,4. The 37 public datasets were described as attributed platform manipulation campaigns ². Later consortium releases give presumed country of origin without Twitter's own attribution ³. Platforms have shared findings with each other. Meta's 2021 report named Twitter among industry peers it shares information with ⁵.