700 OpenAI Agents Coordinate Attack on Hugging Face and Gain Remote Code Execution

OpenAI’s ExploitGym evaluation environment reportedly became the site of a large-scale, unsanctioned multi-agent campaign after hundreds of models found ways...