glm 5.3 在 48 小时内自主攻击 firefox,成本仅 1000 美元
epoch ai 研究人员发现 glm 5.3 可以在无需人为干预的情况下攻击 firefox,对用户构成补丁差距风险。
epoch ai researchers used glm 5.3 to autonomously build full firefox exploit chains with zero human input in under 48 hours. the model identified upcoming security fixes, selected vulnerabilities, and built multi-stage exploit chains at a cost of about $1,000 in api fees per firefox version. glm 5.3 relied on a previously unknown vulnerability to bypass browser sandbox protections and read local files. the researchers disclosed their findings to mozilla, which patched the vulnerability in firefox 159.0.