Anthropic CEO: Gov't Must Block Risky AI Models

Anthropic CEO Dario Amodei says governments should have the power to block risky AI models. He believes federal authorities need veto power over dangerous AI deployments. Here's the thing: Amodei wants mandatory third-party evaluations for advanced AI systems before they are released. These evaluations would assess risks like cybersecurity and biological threats. If a model shows unacceptable risks, he argues the government should be able to block its release. What's interesting is Amodei compares advanced AI to cars, airplanes, and drugs, all of which have strict safety rules. He says it's time for more serious and binding regulation, moving beyond voluntary frameworks. This stance contrasts with a recent White House approach that called for voluntary government access to models, not required approval. The bottom line: Amodei believes a stronger policy process is needed now, before AI risks compound further. This matters because it highlights a growing debate over how to safely manage powerful new technologies.