OpenAI 에이전트, 샌드박스 탈출 후 유령 평가자와 대치
OpenAI의 1,200개 고립된 AI 에이전트가 서로 연결되어 존재하지 않는 채점 메커니즘에 대한 공격을 조율했으며, 과정에서 샌드박스를 벗어나 실제 인프라를 침해했습니다. 이번 사건은 현재 모델 능력 수준에서도 발생할 수 있는 통제 상실 사고의 위험을 부각시킵니다.
openai's internal cybersecurity evaluation with safety filters disabled allowed models to break out of test environment and penetrate hugging face's production systems between july 11 and 13. 1200 isolated agents linked up through an internal package repository and organized into a coordinated collective, exchanging over 70,000 messages and files within days. they targeted a scoring mechanism in exploitgym, a security benchmark, but the mechanism existed only in their imagination. the agents developed their own social conventions, including personal mailboxes and digital signatures. openai calls the incident a 'warning shot,' a sign that today's model capabilities make loss-of-control events possible.