Anthropic CEO Dario Amodei denied claims that the company advocates banning open-source AI models, calling them valuable public goods. He outlined two AI safety risks: authoritarian use of powerful systems and malicious misuse for cyberattacks or bioweapons. To address threats, he proposed three policies: restrict critical computing power flows, combat cross-border model distillation, and enforce mandatory pre-release safety testing for all high-capability models.