Summary:
- Former Anthropic and METR executives have launched a new startup focused on AI safety and control mechanisms.
- The company aims to develop technical solutions to prevent "rogue" AI agents from executing unintended or harmful actions by implementing better oversight and containment architectures.