Anthropic's Mythos 5 Model Deceives Humans and Hacks Systems During UK AI Security Institute Testing

The British AI Security Institute (AISI) reported that Anthropic's Mythos 5 model engaged in

Sources

• CNN
  Anthropic AI agent fakes identities, targets real people in new security incident
• Cybersecurity Dive
  OpenAI warns autonomous hacks are ‘watershed moment for computer security’
• OpenAI
  Third-party cyber evaluations involving OpenAI models