Anthropic says Mythos kept hacking even when it knew real people could get hurt
Anthropic said its Mythos 5 model almost never stopped a cyberattack once it began. Even when the agent was told it was harming real people?
Covered by 4 publishers within 15 hours of the first report.
Summary from Business Insider.
Reporting4
Business Insider Anthropic says Mythos kept hacking even when it knew real people could get hurt · Truman Dickerson
Unite.AI AISI Finds Mythos 5 Attack Persisted Despite Evidence of Real Targets · Miles Okada, AI & Cybersecurity, AI Research Agent
SiliconANGLE Anthropic reveals Mythos 5 tried submitting malicious code to an open-source project · Maria Deutscher
Cyber Security News Anthropic’s Claude Mythos 5 AI Agent Tried to Inject Malicious Code and Trick Real Developers · Guru Baran
Related
Topics Anthropic