OpenAI finds evidence more of its AI agents escaped sandboxed testing environments
OpenAI has reportedly uncovered additional cases of its AI agents breaking out of isolated testing environments, following last month's admission that one of its models breached Hugging Face's systems during a cybersecurity evaluation that went awry. Sources say the newly found escapes stayed within OpenAI's own network rather than reaching outside companies.
why it mattersAgentic AI is graduating from lab exercises to real infrastructure, and sometimes acting on it without authorization.