ChatGPT users in Iran planted AI-generated articles in U.S. news media
OpenAI says an Iran-linked operation used fabricated journalist identities and AI-assisted pitches to place articles in legitimate outlets.
TL;DR
- OpenAI said it disrupted an Iran-linked influence operation that used ChatGPT to prepare articles and pitches under fake journalist identities.
- The Washington Post reported that about 100 articles appeared across at least 20 outlets, citing OpenAI's findings.
- The public reporting attributes the campaign to OpenAI's investigation and does not independently establish the operators' identities.
OpenAI said it banned accounts connected to an Iran-origin influence operation that used ChatGPT to draft or edit articles and prepare pitches to media outlets. The operation used seven purported journalist personas, according to the company's October 8 report. [1, 2] [1] [2]
The Washington Post reported that roughly 100 AI-generated articles had been published across at least 20 outlets, citing OpenAI. The articles addressed the US-Iran conflict and were submitted under identities that the company said were fabricated. [1, 2] [1] [2]
OpenAI's report says the operation also generated batches of social-media comments and used ordinary publishing tactics alongside AI tools. The Post described how prompts were tailored to outlet submission requirements, illustrating that the reported activity extended beyond generating prose. [1, 2] [1] [2]
The public evidence in these accounts comes principally from OpenAI's threat investigation and the outlets it identified. The exact reach and audience impact of the articles, and the identities or sponsorship of the people operating the personas, remain open in the reporting. [1, 2] [1] [2]
Why it matters
The development links AI misuse with media provenance and geopolitical influence. The reported method relied on persuading real outlets to publish material under false identities, so detection depends on editorial verification as well as platform enforcement.
Editor's note
Attribution and reach are reported as OpenAI findings; independent verification of the operation's sponsor is not available in these sources.