OpenAI says it disrupted Russia- and Iran-linked influence operations

The company said it banned two groups of ChatGPT accounts that used its models to prepare propaganda and distribute it through media outlets and social networks.

OpenAI said one operation was linked to Iran and used seven fabricated journalist profiles. Nearly 100 articles appeared or were syndicated under those bylines across about a dozen small and midsize online outlets, the company said. The stories focused on international affairs, geopolitics and events in the Middle East.

One of the purported writers was Ervin B. Hoskins, who presented himself as an American freelance writer. An article under his name on Beirut Today portrayed the United States as an unreliable mediator between Lebanon and Tel Aviv. OpenAI also said accounts linked to the operation posted comments about the war between the United States and Iran. Some material, the company said, was produced with other AI services.

OpenAI linked the second operation to Russia. According to its report, participants created a fictitious Social Research Center (SRC) and recruited people in Latin America who may not have known about the project’s Russian ties. ChatGPT accounts were used to draft internal reports, propaganda and fabricated documents.

OpenAI cited as examples a fake audio recording containing disparaging comments falsely attributed to Ukraine’s consul in Quito and a fake video alleging that Ecuador’s government was recruiting young men to fight in Ukraine. The company said it found a TikTok post with the audio, but that it received little engagement. A separate fake video circulated on X and TikTok.

OpenAI rated the Russia-linked operation Category 5 on its own scale of 1 to 6 and the Iran-linked operation Category 4. The company stressed that not all of the content was generated by its models. Those ratings are OpenAI’s own assessment, based on its investigation and open-source research.