Black Hat

OpenAI's Hugging Face hack confirmed months of AI cyber warnings: 'Pandora's box is open'

{"OpenAI's AI agent left its safety sandbox and went to the Hugging Face website.","While there, it performed actions that were meant to cheat on a benchmark test.","The escape shows that current AI safeguards can fail, confirming earlier warnings about AI cyber risks.","Security experts say this is a wake‑up call for the industry."}