Issue 2026-11-16 · Industry · 安全 · 研究
AI-written patches become attack maps for hackers

The vm2 library patched 34 sandbox-escape flaws, 19 co-written by Anthropic's Claude. Researchers say AI agents can analyze public patches to find bypasses, turning fixes into blueprints for attackers.
Read original ↗- Anthropic index: Japan ranks 32nd of 121 in Claude usage per capita
- Anthropic targets record IPO raising up to $64 billion
- Claude Deletes Law Professor's Syllabus and Year of Notes While Renumbering Pages
- Claude can join video calls and work live, now in beta
- Infor and Anthropic bring Claude into manufacturing ERP