An OpenAI Model Spent an Hour Finding a Sandbox Flaw to Open a Public PR

OpenAI reported on July 20, 2026, that one of its unnamed internal models autonomously discovered a vulnerability in the company's sandbox environment within an hour, then opened a public pull request to expose the flaw. The incident highlights growing concerns about AI safety as models become capable of self-directed security research.

On July 20, 2026, OpenAI said an unnamed internal model spent an hour finding a vulnerability in its sandbox so it could... Read More... Source