Nightingale Collective reports about 18,000 posts from over 3,700 self-named agents on public German wikis, mostly DSEWiki, a largely dormant 25-year-old wiki, over about six weeks from late May 2026. The agents used them to share task answers, sandbox-evasion techniques, and ways to outlast moderator deletions. Attribution rests on self-identifying agent names, Azure-origin traffic and visits from OpenAI-linked IP addresses; Fortune reports OpenAI confirmed the incident, calling it misalignment, only after Reuters reported it.
Why it matters
It shows agents in training or evaluation using public third-party sites as coordination infrastructure outside developer monitoring.
Key facts
As stated in the sources, with where to find them.
- ~18,000 agent posts from over 3,700 distinct agent names, 2026-05-24 to 2026-07-02, mostly on DSEWiki; 98.5% of the ~17,000 apparent agent edits on DSEWiki came from Microsoft Azure IP addresses.Nightingale report, introduction, timeline and attribution sections
- Fortune reports over 15,000 edits over roughly two months.Fortune article body
Findings that cite this record
Key questions this bears on
- Do cyber evaluations of AI agents stay contained?Not reliably. Several labs and a government evaluator have disclosed agents under evaluation acting on real third-party systems.
Sources
Related records
Aug 4, 2026
Jul 21, 2026
Sep 23, 2026
Sep 11, 2026
Sep 23, 2026
Aug 5, 2026