OpenAI Cuts Off Russian Accounts Using ChatGPT to Fake Political Voices Online
A network hiding behind VPNs used the chatbot to churn out posts for a phony think tank called the International Burke Institute across Substack, Telegram, X and LinkedIn.
Key points
- OpenAI on Tuesday banned a cluster of Russian ChatGPT accounts running an influence operation, meaning a coordinated campaign to sway public opinion with fake voices.
- The accounts used VPNs (virtual private networks, tools that hide a user's real location) to get around OpenAI's regional access rules.
- ChatGPT wrote social media posts pushed onto Substack, Telegram, X, Facebook and LinkedIn.
- The content promoted a front group calling itself the International Burke Institute (IBI).
- OpenAI disclosed the takedown in its latest threat report, which tracks abuse of its tools.
OpenAI has pulled the plug on a batch of Russian-run ChatGPT accounts that were quietly using the chatbot as a propaganda factory, with Tuesday's announcement landing a day after we reported OpenAI cutting off a Cambodia-based scam ring doing similar work through ChatGPT.
They weren't writing posts by hand. Operators fed prompts into ChatGPT and let the AI produce the words, which then landed on Substack, Telegram, X, Facebook and LinkedIn.
Who was behind the accounts?
OpenAI attributes the activity to a Russia-linked network. To reach ChatGPT, the operators used VPNs, software that reroutes internet traffic through another country to disguise a user's real location. OpenAI restricts access from certain regions and the VPNs were the workaround.
The report, first surfaced by The Hacker News, describes the cluster as one of several state-aligned abuse cases OpenAI has caught this year.
What were they actually posting?
The content pushed a group calling itself the International Burke Institute, or IBI. On paper it looks like a policy think tank. In practice, according to OpenAI, it functioned as a front: a respectable-sounding label to wrap political messaging so posts looked like commentary from a real research body rather than a coordinated campaign.
That's the appeal of generative AI for this kind of work. One person can produce the writing output of a small newsroom, in multiple languages, without hiring anyone.
How did OpenAI spot it?
OpenAI's internal threat-intelligence team looks for patterns of misuse across accounts: shared prompts, shared writing styles, shared infrastructure, odd log-in behaviour. Once a cluster is confirmed, the accounts are terminated and findings go into a public threat report.
The company has published these reports roughly every few months since 2024. Earlier editions named operations linked to Russia, China and Iran, among others.
Should ordinary social media users care?
Yes, though not in a panicked way. A share-worthy political post from an account you've never heard of, citing a think tank you've never heard of, deserves a second look before you repost it.
Two habits help. Check whether the "institute" or "foundation" behind a viral claim has a real staff page and real funding disclosure. Cross-check any explosive claim against an established news outlet before sharing.
None of that is new advice. AI-generated propaganda just makes it cheaper to produce, so the volume goes up, and that's the part worth watching.
What OpenAI has not said
The company didn't publish the exact number of banned accounts, nor the reach of the posts in terms of follower counts or engagement. It also didn't name individuals behind the IBI branding. Attribution in influence operations is hard, and vendors tend to stop short of naming people without hard evidence.
Expect more of these takedowns. The economics of AI-assisted propaganda favour the attacker, and platforms on both sides of the problem know it.
