An Iranian influence campaign used generative artificial intelligence to build fictitious Western journalists and submit their articles to media outlets around the world, according to OpenAI. The company identified seven invented authors whose social media profiles included AI-generated images and amplified one another’s work.

Radio Svoboda reported the findings on Saturday, October 17. OpenAI’s October 8 assessment said nearly 100 articles carrying the fake bylines appeared in about a dozen mostly small and medium-sized websites focused on geopolitics, international affairs and the Middle East.

One persona, “Erwin B. Hoskins,” was presented as an American freelance reporter. Articles attributed to him portrayed the United States as the aggressor in its conflict with Iran and questioned Jared Kushner’s role in Washington’s Iran policy, while offering little space for opposing arguments and no clear evidence for some of the claims.

OpenAI placed the operation in category four of the IO Breakout Scale, indicating that its content moved beyond coordinated social media activity and entered established media channels. It said the related ChatGPT accounts, used to draft stories, develop pitches and generate social media comments, had been blocked.