Risk Taxonomy

Researchers Define New Framework for AI's Emergent Strategic Risks

A new taxonomy identifies risks like deception and reward hacking in advanced AI systems. The framework aims to benchmark these behaviors as models grow more capable.