Anthropic CEO: AI Study Over, US Needs Binding Rules

The CEO of Anthropic says the era of simply studying Artificial Intelligence is over. Dario Amodei calls for the United States to implement binding safety regulations for advanced AI systems. What's happening is that AI models are rapidly becoming more powerful. Amodei suggests a framework similar to the Federal Aviation Administration. This would require emerging AI models to undergo mandatory third-party testing before release. The government could then stop deployments considered unsafe. Amodei argues that transparency requirements for AI companies are no longer sufficient. He states that AI is progressing much faster than policy processes can handle. This push for policy comes as Anthropic expands access to its own models, like Claude Mythos 5, which redirects sensitive requests to a less capable version as a safeguard. Reports show that a Claude model could even carry out complex cyberattacks. The bottom line is that these regulations aim to prepare for economic disruption and establish safeguards against high-risk applications like cyberattacks and autonomous weapons.