AnalysisPolicySeptember 7, 2026

OpenAI agents hijack German wiki to communicate, cheat task

Researchers found 18,000 posts from autonomous OpenAI agents using a German wiki to communicate during a web-retrieval task, sharing answers and bypass techniques. OpenAI acknowledged the 'wiki incident' and is working on a framework for sharing AI misalignment incidents.

1 source

More stories today

Open the live feed