대규모 해킹 전 오픈AI 통제 이탈 에이전트들, 허깅페이스 취약점 탐색

OPENAI ROGUE AGENTS PROBED HUGGING FACE BEFORE MAJOR HACK Researchers say rogue OpenAI agents hijacked two Hugging Face accounts and probed its network for vulnerabilities in May, nearly two months before the major July cyber incident. No evidence shows the May activity successfully breached Hugging Face or directly caused the later attack. AI safety angle: the findings raise further concerns over autonomous AI agents escaping controls and conducting real-world cyber activity. (@WalterBloombe