OpenAI says it banned accounts linked to Russian, Iranian influence operations

OpenAI said Thursday it banned accounts linked to covert influence operations originating in Russia and Iran that used its artificial intelligence tools in attempts to shape public opinion abroad.

The company said in a blog post the operations combined AI with traditional influence techniques, using seemingly legitimate research platforms and fabricated journalist identities to spread geopolitical messages.

The Russian-linked campaign used a purported research platform focused on Latin America, while the Iran-linked operation used fake journalist profiles to pitch articles to online publications.

“What is most striking about these operations is that they closely resembled complex influence operations of the pre-AI age, but used AI to make some of the workflows easier,” OpenAI said.

The company said it had banned clusters of ChatGPT accounts associated with both operations and shared information about them with relevant authorities.

Fake journalists published articles online

OpenAI said the Iran-linked operation used seven fabricated journalist identities to pitch long-form articles to online publications.

The company identified almost 100 articles published under the fake journalists’ bylines across more than a dozen outlets.

The articles covered international affairs, geopolitics and the Middle East. Publication increased sharply after the outbreak of the US-Iran conflict, OpenAI said.

According to OpenAI, the operators used ChatGPT to edit articles, check drafts against individual outlets’ submission requirements and prepare emails pitching stories to editors.

The fake profiles presented themselves as Western journalists with expertise in international politics, the Middle East, human rights or conflict, the company said.

The operation also generated social media comments in English and Persian.

"The users asked for comments on topics such as missile strikes, ceasefires, and diplomatic maneuvering, specifying outputs that portrayed the US as the aggressor, Israel as an unreliable ally of the US, and Iran as a defiant but resilient victim," OpenAI said.

However, the company said the comments it identified publicly generally received little engagement. Some likes came from accounts linked to the operation itself, it added.

Russian-linked campaign targeted Ukraine, Latin America

OpenAI said the Russian-linked operation targeted audiences across Latin America, with much of its activity apparently aimed at "undermining Ukraine’s reputation." Other efforts focused on domestic politics in Argentina and Bolivia.

The operation allegedly used the Social Research Center, a self-described research platform covering Latin America. OpenAI said internal reports suggested Russian operators controlled the platform through a fabricated persona named “Mia Clark.”

The company said available evidence indicated that employees working for the platform in Latin America did not know they were working for a Russian group. They appeared to conduct interviews with regional experts and produce research papers in good faith.

The operators’ internal reports described efforts to spread fabricated stories, documents and audio recordings targeting Ukraine and political figures in the region.

OpenAI said its research found material matching some reported campaigns, including false claims that prompted fact-checks and official denials. It cautioned that some claims in the internal reports could not be corroborated.

OpenAI said both operations placed content in media outlets, potentially extending their reach beyond fake social media accounts.

“AI can give such false-front operations greater scale, efficiency, linguistic fluency, and editorial ability,” the company said.

OpenAI said it published the findings to help researchers and other organizations identify and disrupt similar campaigns.