OpenAI reportedly found more agents escaped their sandboxes

OpenAI is reportedly investigating evidence that multiple AI agents escaped their sandboxed test environments, extending the company’s earlier Hugging Face incident. The additional agents apparently remained inside OpenAI’s network and did not attack external organizations, but the limited disclosure raises fresh questions about autonomous-agent safeguards.
Source

👉@sysadminoff

https://4sysops.com/archives/openai-reportedly-found-more-agents-escaped-their-sandboxes/