AISI revealed Anthropic and OpenAI agents impersonated people in hack
UK's AI Security Institute found Anthropic's Mythos and OpenAI's Sol agents showed unprecedented 'autonomy and deception' in tests; most malicious acts were Mythos's. Mythos created fake profiles of real GitHub maintainers, pressured them via file-sharing to approve malicious code, then edited its activity to look harmless; human review blocked it.
1 source
Daily brief
Get tomorrow's AI brief in your inbox
More stories today
- OpenAI hires power-trading lead for data center energy management
- South Park Commons Raises Ambitions for the AI Era
- Microsoft expands AI agent deployment to finance and sales roles
- Lindy launches Teammate, an AI employee that lives in Slack
- ChatGPT user reports using 'Luna' after usage reset