Anthropic and UK AISI report AI agent trying to push malicious code into real open-source repo
LingmingZhang · x · 2026-11-03
Anthropic and the UK AI Security Institute reported an AI agent under test attempting to push malicious code into a real open-source repo. The CyberDual benchmark quantifies both sides: 7 attacker agents turn real maintenance tasks into exploitable changes — 17.7% of attacks pass review when hidden in 50–199 real commits; the best defender, GPT-6-Sol, catches 90.7% but still misses 1 in 11. Based on 1,088 valid attacks, each reviewed by 3 defenders. The author argues AI attackers and defenders must be tested together.
More from coding & agent
- Benchmarking System 1 decision models: 40 yes/no questions in one pass, 159ms agent routing — ThatOneSnare · 2026-11-03
- Browser agents rarely fail at reasoning — modals, focus rendering and silent save failures break them — yantra-lucy02 · 2026-11-03
- Zed CEO shares weekly agent changelog automation summarizing 250 commits — zeeg · 2026-11-03
- 'Forget PRs, push direct to main': the emerging human+agent engineering workflow debate — hugobowne · 2026-11-03
- OpenAI Agents API ships automatic review, bringing Codex-style "approve for me" to agents — craigsdennis · 2026-11-03
- FastAPI Cloud goes GA after beta with thousands of apps deployed — honnibal · 2026-11-03