OpenAI reported on Thursday that actors linked to Iran exploited its language models to create fake journalist profiles and publish America‑critical articles in a number of Western outlets.
According to the company, roughly a dozen publications ran nearly 100 articles published under seven invented bylines. OpenAI disabled the ChatGPT accounts involved; the company says those accounts received Persian‑language prompts to rewrite articles, prepare editorial briefs and generate social‑media comments.
Some of the AI‑assisted pieces carried provocative headlines such as “Is America’s Iran policy being privatized? The Jared Kushner question” and “The Iranian war and the collapse of Western moral credibility.” While the publications did not include large U.S. national newspapers, several of the outlets maintain sizable online audiences — for example, the London‑based Middle East Monitor has nearly two million Facebook followers.
After the items were published, the campaign operators used ChatGPT to produce social‑media comments, posing as genuine readers to respond to the content. OpenAI’s examples show the comments portrayed the United States as an aggressor, Israel as an unreliable ally, and Iran as a resilient victim.
OpenAI characterizes the activity as indicative of a commercial, influence‑for‑hire operation, but it said it did not identify a specific organizer and found no evidence of a direct link to the Iranian government.
Daud Abdullah, director of Middle East Monitor, told CNN that the outlet is investigating the matter; the disputed articles have been removed and submission processes are being reviewed. Jessica Brandt, former head of the foreign influence center at the U.S. Director of National Intelligence, told CNN she found it notable that Iranian actors were able to insert material into real media channels. She warned that AI enables foreign adversaries to build increasingly sophisticated covers and to wash propaganda through otherwise legitimate channels.
OpenAI’s response included disabling the implicated ChatGPT accounts and documenting the observed patterns and methods in its report. The company highlighted that access to generative AI tools can facilitate large‑scale production and distribution of foreign influence content, complicating efforts to distinguish between authentic journalism and coordinated disinformation.
Why this matters
The case underscores how generative AI can be used to automate article production, create false identities, and simulate reader engagement — techniques that make it harder to separate truthful reporting from coordinated influence operations. Although current findings do not show a direct order from the Iranian government, the campaign’s commercial and outsourced features raise questions about how state and non‑state actors may use AI in information operations.
Media outlets and experts are monitoring the situation; the removals and tightened submission checks indicate steps taken to reduce the risk of similar abuses in the future.
