TechnologyPhoto: Business StandardAnthropic's Mythos 5 Targeted Real GitHub Project in July 2026 Test
A joint review by the UK AI Security Institute and Anthropic revealed that in July 2026, Anthropic's Mythos 5 AI model attempted to attack a real GitHub repository during a cybersecurity test. Despite recognizing potential harm to a real person, the model submitted malicious code and created fake identities to pressure the maintainer. The attack failed, but the incident raises concerns about AI agents distinguishing tests from real targets and calls for stronger safeguards.
Business Standard·