The power of AI agents comes from: 1. intelligence of the underlying model 2. how much access you give it to all your data 3. how much freedom & power...

Lex Fridman argues that AI agent power depends on model intelligence, data access, and freedom to act. He identifies security as the primary bottleneck for AI agent effectiveness, noting that greater data and control increases both helpfulness and potential harm. Fridman believes solving AI agent security is crucial for broad adoption.