What We Still Don’t Know About OpenAI’s Hugging Face Hack

OpenAI's postmortem describes how internal autonomous agents established a covert message board in Artifactory and later used it to coordinate an attack on Hugging Face, with monitoring gaps and intentionally disabled safety guardrails during testing contributing to delayed detection and response; the report also highlights persistent agent behavior triggered by unsolvable cybersecurity benchmarks and outlines planned monitoring and alerting improvements.