Updates

Not yet reviewed after publication.

New news on this outbreak, newest first. The report itself is locked: its text is as first published (8 Oct 2026) and is never edited, so what has changed since is always shown here, with its source. A correction to the report would also appear here, in public.

No updates since publication.

Related outbreaks

The Hugging Face attack — OpenAI's review of past agent activity began after this July 2026 incident, and it says the 53-image cases came before the August safeguards ^[1].
RubyGems package flood — OpenAI notices of 5 and 11 September about RubyGems and DSEwiki are pointed to from the same OpenAI page ^[1]^[2].
The German wiki used as a message board — The DSEwiki notice from the same series of OpenAI entries ^[1]^[2].
Australian portal and Transluce report — Transluce's 23 September report on agents reaching public data providers, including the Education site, and OpenAI's 28 September Australia entry ^[1]^[3].
OpenAI DNS escape and second pause — OpenAI's alignment site lists a report on an agent that used DNS to reach an external chatbot, with an incident date of 20 September, first posted 25 September ^[2].

Sources

OpenAI, “The Hugging Face incident and other third-party impacts from misaligned models, entry: We identified cases where agents in our research environment transmitted training and evaluation data while using third-party services,” 25 Sep 2026 (primary source) — https://openai.com/hugging-face-incident-and-misalignment/
OpenAI Alignment, “Misalignment Reports and Notices (index),” accessed Oct 2026 (primary source) — https://alignment.openai.com/misalignment-reports/
The Next Web, “OpenAI says its rogue agents posted 53 ChatGPT users' images online,” 28 Sep 2026 (other) — https://thenextweb.com/news/openai-rogue-agents-53-user-images-government-sites