OpenAI identified and blocked an influence operation, named Bogus Bylines, which utilized its AI models to generate content. The campaign involved the creation of seven fake journalist personas to pitch and publish articles in various media outlets.
Between July 2025 and October 2026, approximately 100 articles were published in publications including Daily Kos and Middle East Monitor. The content focused on criticizing US military actions in Iran.
Operators used Persian-language prompts to generate English-language text via ChatGPT and employed VPNs to bypass existing regional access restrictions. Additionally, the group used the AI to draft editorial pitches and fabricate social media comments.
OpenAI assigned the operation a Category 4 rating on its six-level influence scale, noting that the content reached millions of readers. The company banned the associated accounts on October 8, 2026.
