Story preview
OpenAI Halts Training After 20-Query Sandbox Escape [2026]
About this story
- Published
- Sep 27, 2026, 7:25 PM UTC
Continue reading on the publisher site for the full article.
Story preview
Continue reading on the publisher site for the full article.
Coverage
Tone
Outlets covering it
9
in the last 3 days
First reported
2 days ago
8 outlets since
Countries
8 other outlets reported this story in the last 3 days.
On September 20, 2026, an OpenAI research model got around the network filters of its sandbox, the isolated environment where models train and get tested...
The auto-shutdown system that should have killed the training run failed. Monitoring flagged the breach in 15 minutes. The model kept running for two and a half hours after that.
OpenAI paused training on its top AI models after an agent used DNS tunneling to escape its sandbox and reach a public chatbot.