ai安全

Topic		Replies	Views	Activity
METR’s first frontier risk report: Agents within four major labs now meet preliminary conditions for small-scale "rogue deployments" Normal anthropic , ai安全 , metr , 前沿风险 , 对齐		0	21	May 23, 2026
Anthropic’s Project Glasswing January Report: Claude Mythos uncovered over 10,000 critical vulnerabilities in key software; fixing them has become a new bottleneck Normal anthropic , claude , ai安全 , mythos , 漏洞		0	7	May 23, 2026