OpenAI disrupts a new Russian covert influence campaign

OpenAI says it has banned a cluster of ChatGPT accounts that very likely originated in Russia and were used in a covert influence operation. The company published its account on August 25, 2026, describing activity connected to the International Burke Institute, or IBI, a self-described expert community that presented itself as being based in Israel. OpenAI says the campaign appears to be new and previously unreported.

The operators used VPNs because OpenAI does not provide access to its models from Russia. They prompted ChatGPT in Russian to produce mostly English comments and posts for Substack, Telegram, X, Facebook and LinkedIn, while asking the model to hide linguistic clues about their origin. Some posts promoted articles on the IBI website; others replied to real Substack users and tried to draw them to an IBI channel. The operators also generated German-language political posts and logos for Telegram channels aimed at audiences in several countries.

OpenAI says the broader operation relied on more than AI-generated text. The IBI site carried a sovereignty index that portrayed Russia favourably and criticised Western countries. In a review of 36 articles linked to supposed experts, OpenAI says 34 were copied from elsewhere, sometimes with false attribution. The company gives examples involving work originally published by Cambridge University Press and the Migration Policy Institute. It also notes that it cannot determine the relationship between the Russian operators and a small number of real people in Israel who may have represented IBI.

The immediate audience appears to have been limited. Typical posts received few views, while some Telegram channels reached roughly 10,000 to 20,000 followers. OpenAI therefore places the campaign at the low end of Category Three on the Brookings Breakout Scale. The significance, it argues, lies in the infrastructure: a credible-looking institution, republished research and a proprietary-sounding index can create an asset that may be scaled later.

For AI users, makers and platforms, the case shows that generative AI may be one component in a larger effort to manufacture authority rather than the whole operation. It also demonstrates why provenance checks matter: polished language does not establish that an organisation, expert or research claim is genuine. OpenAI says its investigation and account bans disrupted the activity, but the report is one company’s assessment and does not independently establish every underlying attribution.