Anthropic CEO Dario Amodei on Thursday backed a testing regime for frontier AI models that would allow governments to block or deter deployment if an independent third-party auditor deems it too risky for public release.
Amodei’s comments, published in a blog post, represent the most aggressive regulatory framework backed by a major AI CEO to date. The proposal follows an executive order on AI oversight signed by President Donald Trump on June 2 that gives the intelligence community an enhanced role in model testing.
Comparing AI oversight to the Federal Aviation Authority’s approach to aircraft safety, Amodei said, “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.”
According to Amodei, private organizations should be empowered to audit models across four main risk areas: “cybersecurity, biological weapons, loss of control of AI systems and automated R&D,” referring to AI models’ potential ability to improve themselves autonomously.
Trump’s earlier executive order set out a government-led voluntary testing protocol under which AI companies are asked to submit powerful new models for review 30 days before public release.
The White House did not immediately respond to a request for comment.
Amodei also wrote that policy responses to address potential AI-driven labor market displacement could include the creation of “universal capital accounts” or other tax vehicles — wading into a thorny political fight that has divided Washington.
Sen. Bernie Sanders (I-Vt.) last week proposed a 50% tax on stock offerings of the top AI companies that would give the government a voting stake in companies’ decisions. Days later, Trump floated an unspecified “partnership” with leading AI firms, which could include disbursing companies’ dividends to Americans, in an effort to secure a broader public buy-in.
Anthropic has considered wealth-sharing frameworks for at least the past few months. In October it published a policy paper that suggested a national sovereign wealth fund with AI investments could help distribute revenue generated by the technology.