According to TechFlow, citing Decrypt on June 11, Anthropic CEO Dario Amodei published an article on Wednesday calling on the U.S. government to impose binding safety requirements on the most powerful AI models. In the article, titled “Policy and AI’s Exponential Progress,” Amodei argued that the era of transparency-only regulation has ended and that frontier AI systems should be subject to a more enforceable safety regime.
Amodei compared the regulation of frontier AI systems to the safety oversight applied to aircraft. In his view, advanced AI systems should undergo technical testing and audits before release, and systems that fail to meet safety standards should be barred from being launched. The proposal shifts the focus from disclosure alone toward direct assessment of model capabilities, risk boundaries, and deployment conditions.
The proposal calls for mandatory third-party testing of advanced AI models across four categories of risk: cybersecurity, biological weapons, loss of control, and automated research and development. It also seeks to give the government authority to prevent unsafe deployments, allowing regulators to intervene when a model has not satisfied required safety standards.
The proposal comes as Anthropic is preparing for an initial public offering. The company previously completed a $6.5 billion Series H financing round, reaching a valuation of $96.5 billion. Amodei’s call for stronger regulation therefore arrives at a moment when Anthropic’s own corporate development and the governance of frontier AI models are both in focus.
The issue remains contested within the industry. Critics, including OpenAI CEO Sam Altman, have questioned whether strict regulation could become a justification for allowing a small number of companies to concentrate control over the technology. The debate reflects differing positions on safety standards, regulatory intensity, and the degree of openness around frontier AI development.