OpenAI bans accounts behind fake journalists and think tank [1]

OpenAI said on October 8, 2026, it banned ChatGPT accounts linked to Russian and Iranian influence campaigns using fake journalists and a purported think tank. According to its findings, the deception reached editors and research staff, giving operators a route to audiences through other people's work.

The Russian campaign, called Dark Clark, apparently controlled the Social Research Center in Latin America through an invented identity, Mia Clark. OpenAI says local employees appeared unaware of the Russian connection, and the accounts mostly used ChatGPT to draft internal reports. The Iranian campaign, Bogus Bylines, used seven invented journalist identities to pitch articles to news outlets, according to OpenAI.

What the reach rating means

OpenAI assigned Dark Clark its first Category 5 rating for a disrupted influence operation. That deserves a careful reading. The Breakout Scale, published by Brookings, classifies campaigns by how their messages spread across platforms and communities, including into media and political debate. Category 4 means mainstream media amplification; Category 5 means amplification by prominent people, such as political candidates.

Those categories describe observable distribution. They shouldn't be read as a count of persuaded readers or proof that a campaign changed an election outcome: that's our interpretation of what the scale measures. OpenAI also warns that operators took credit for unrelated events and exaggerated results.

The next thing to watch is evidence that separates public amplification from actual persuasion, a distinction the scale alone cannot settle. What evidence would you want a newsroom to require before accepting work from an unfamiliar contributor?