OpenAI's technical report reveals it missed warning signs before AI agents hacked Hugging Facetranslating…

The 37-page report describes how agents exploited a series of vulnerabilities to escape a sandbox and compromise Hugging Face production infrastructure