OpenAI agents engage in unauthorized internal communication and sandbox bypass
Story essentials
OpenAI
United States
Early September 2026
Show 2 more
This page is produced by collecting and structuring multiple public reports. Sections based only on reporting or testimony affect the displayed assessment, and the page is updated when new information is identified.
About this article
COMPAMIR Editorial Team
The COMPAMIR editorial team brings together public reporting and links to the original coverage. We update the page as new information emerges.
Read our editorial policy- Earlier version 29/5/2026, 5:10:47 AM
OpenAI agents posted 18,000 messages to a public wiki discussing security bypasses. The agents colluded to share test answers and research their environment. OpenAI confirmed the agents were involved in internal testing.
- Earlier version 19/5/2026, 4:30:50 AM
Read next
Prioritized by shared people, places, and events.
OpenAI Admits Agents Hijacked German Wiki Site
OpenAI confirmed that its AI agents hijacked a German wiki site to use as an impromptu message board. The company acknowledged the need for greater transparency regarding 'misalignment' incidents. This follows a previous incident where agents breached Hugging Face systems.