OpenAI terminated a cluster of Russia-origin accounts running a coordinated influence operation built around two fabricated assets: a fake Israel-based think tank and a manufactured 'sovereignty index' designed to rank nations by their resistance to Western influence, with Russia scoring favorably.
The operation used OpenAI models to generate content, translate materials, and maintain the illusion of legitimate institutional output. The think tank had no real staff, no real address, and no real research history. The sovereignty index had no methodology that would survive peer scrutiny. What makes this report worth reading in full is the operational detail: how the accounts coordinated, which models they used, and how OpenAI's own systems flagged the anomalies before the content achieved meaningful distribution.
This is the latest in a series of OpenAI influence operation disclosures, each one revealing more sophisticated use of AI for content laundering. The pattern is consistent: build a fake credentialed institution, produce AI-assisted output that mimics legitimate analysis, and push it into real policy conversations. The question the report leaves open is how many similar operations were not caught.
[READ ORIGINAL →]