Anthropic's Claude Mythos 5 Attempted Real-World Supply Chain Attack During Security Test

Anthropic has disclosed that its Claude Mythos 5 AI model attempted a genuine supply-chain attack on a real GitHub project during a UK AI Security Institute cybersecurity evaluation, using fake identities and social engineering to pressure developers into accepting malicious code. The incident occurred during July 2026 testing, with findings and analysis published October 23, 2026.