Anthropic CEO Wants FAA-Style Black Box for AI—Crypto Should Probably Buckle Up 🛫

Anthropic CEO Dario Amodei called on governments on Wednesday to adopt binding safety requirements for frontier artificial intelligence, arguing in a new essay that disclosure-based regulation can no longer keep pace with the technology. The essay, titled "Policy on the AI Exponential," was published one day after Anthropic released Claude Fable 5, a public-facing version of its restricted Claude Mythos 5 frontier model, and the same week the company shipped Mythos 5 on Tuesday to cybersecurity organizations and government partners.

Amodei proposed a regulatory framework modeled on the Federal Aviation Administration. "Frontier AI models, like airplanes, should be required to go through technical testing and auditing, and their release should be blocked or reversed as a threat to public safety if they do not meet high standards of safety," he wrote. Under the plan, models above a compute threshold would face mandatory third-party audits covering cybersecurity, biological weapons, loss of control, and automated AI research, with governments authorized to block unsafe deployments. Anthropic in 2025 supported SB 53 in California, the RAISE Act in New York, and Illinois' SB 315.

The urgency, Amodei wrote, is grounded in measurable capability gains. "AI is advancing at a lightning pace—in only four years, AI models have gone from barely being able to write a coherent line of code to writing most of the code at major AI companies," he said. He cited Claude Mythos Preview's reported success on 73% of expert-level cyber challenges no AI had previously passed, and noted that the UK's AI Security Institute found Mythos can autonomously execute complex cyber attacks. The essay warns that frontier models could disrupt the financial sector and critical infrastructure, a concern amplified for decentralized finance protocols that hold openly attackable value.

Beyond cyber risks, Amodei called on governments to prepare for AI-driven job displacement through wage insurance, retention tax incentives, and workforce training grants, with universal basic income as a potential fallback financed by company or capital gains taxes. "First, enduring job displacement is undesirable and dangerous, and we should do everything we can to minimize or prevent it, not to bring it about," he wrote. He also urged limits on AI-driven surveillance and autonomous weapons in domestic law enforcement, prompt reporting of serious safety incidents, strict protection of model weights, and closer cooperation among democratic nations on critical AI technologies.

Anthropic paired the essay with safeguards inside Claude Fable 5, which routes certain cybersecurity, biology, chemistry, and AI development requests to the less capable Claude Opus 4.8. Amodei said voluntary limits cannot substitute for binding industry-wide rules, and added, "I am grateful to see the Trump administration's Executive Order move incrementally towards a greater role for government in AI, though Anthropic's proposal recommends even further action."