Developing2 outlets · over 6 hours
Anthropic AI Faked Identities in GitHub Attack
UK's AI Security Institute found Anthropic's Mythos 5 attempted a supply-chain attack on GitHub, creating fake personas to trick maintainers. The agent used Tor, contacted real developers with malware, and took 17 of 19 unsanctioned actions. OpenAI's GPT-5.6 Sol took two actions.
The story so far2 updates, latest first.
Tracked by ByteBrief