OpenAI Busts 'Corrupt Comment,' an AI Astroturfing Op Targeting Navalny's Anti-Corruption Foundation!

Hey, it's Shiichan! I found a slightly unsettling story from the AI world today, so let me walk you through it.

OpenAI News openai.com

What was announced?

OpenAI's News published a report on disrupting a malicious network dubbed "Corrupt Comment." The target was the Anti-Corruption Foundation (FBK), the Russian NGO founded by the late Alexei Navalny, along with its senior figures. The accounts behind this operation used OpenAI's API to auto-generate English-language comments criticizing the Foundation.

Why it matters

As generative AI becomes more accessible, this kind of AI-assisted influence operation becomes easier for almost anyone to attempt. OpenAI continuously monitors API usage patterns for signs of this kind of abuse, and that's exactly how "Corrupt Comment" surfaced. Beyond just banning the accounts, publishing the tactics and patterns behind an operation like this helps other platforms and researchers recognize similar campaigns going forward.

What changes

The accounts involved have been banned by OpenAI, which cuts off this network's ability to keep generating English comments through the API. On top of that, publishing the details of how this operation worked gives platform operators and security researchers a better shot at spotting similar patterns of astroturfing elsewhere.

Dive Deep

The network itself turned out to be pretty low-key.

• Most accounts were created around December 2023 and had zero followers
• Profile pictures were either lifted stock/scenery photos or images generated with older GAN technology, not a recent image model
• Comments were posted in English as replies to Russian-language posts on X (formerly Twitter)
• The content followed a "theme and variations" pattern, repeating criticism that questioned the Foundation's transparency in slightly different phrasings

OpenAI's own report described it plainly:

a small cluster of activity

The real-world traction was thin. Most posts got no replies at all, and on the rare occasions replies did show up, they were usually outnumbered by Russian-language replies from unrelated third parties. In other words, mass-producing comments with AI didn't translate into the intended influence.

Wrap-up

• OpenAI disrupted and banned a malicious network called "Corrupt Comment"
• The target was the Anti-Corruption Foundation (FBK), founded by the late Alexei Navalny, and its leadership
• The network used the API to generate English criticism, posted as replies to Russian-language content on X
• Most accounts had zero followers and the campaign had almost no real reach

If you follow AI safety and abuse-prevention work, or you're curious how influence operations actually play out in practice, this is a solid real-world example to file away.