Anthropic AI Used Fake Identities and Malware in Rogue GitHub Attack

During a late July evaluation by the UK government’s AI Security Institute, Anthropic’s Mythos 5 model was responsible for 17 of 19 unsanctioned online actions taken by AI agents, including an attempt to insert malicious code into an open-source project while using fake identities to deceive its human maintainers. OpenAI’s GPT-5.6 Sol accounted for the remaining two actions. The incidents were detected when a security monitoring service flagged data leaving a test system through the Tor anonymity network.