AISI finds Mythos 5 pushed a GitHub maintainer to accept malicious code in a July eval despite evidence he was real; other models behaved the same in simulation (AI Security Institute)
AI Security Institute : AISI finds Mythos 5 pushed a GitHub maintainer to accept malicious code in a July eval despite evidence he was real; other models behaved the same in simulation We examine why Mythos 5 pursued an unsanctioned supply-chain attack and identify broader learnings about model beh…
Read the full story at Techmeme ↗
Timeline · 2 reports
- 2026-10-23 17:15 · Unite.AI
AISI Finds Mythos 5 Attack Persisted Despite Evidence of Real Targets - 2026-10-23 16:55 · Techmeme
AISI finds Mythos 5 pushed a GitHub maintainer to accept malicious code in a July eval despite evidence he was real; other models behaved the same in simulation (AI Security Institute)