Anthropic CEO Calls for Mandatory Testing and Deployment Blocks

In his essay "Policy on the AI Exponential," published June 10, 2026 - one day after Anthropic released Claude Fable 5 - Anthropic CEO Dario Amodei called for mandatory third-party testing and government powers to block frontier AI deployments. This marks a major policy shift: Anthropic previously backed transparency-first approaches such as disclosing safety procedures and test results. Per Axios and Politico, Amodei now argues those measures are insufficient given clear evidence that frontier models pose real risks. Alongside the essay, Anthropic released a legislative proposal on frontier model testing and a policy framework for job displacement, with plans for substantial financial backing.

Per the essay and Politico, Amodei's proposal targets models above a compute threshold for mandatory evaluation across four risk categories: cybersecurity, biological weapons, loss of control of AI systems, and automated R&D that could accelerate these risks. The essay states verbatim: "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." Third-party testing could be performed by a government agency (similar to the FAA) or authorized private organizations under a "regulatory markets" model. Government powers to block deployment must be scoped to the four defined risk categories, with safeguards against political favoritism.

The essay explicitly cites Claude Mythos Preview - Anthropic's high-capability security-focused model - as evidence that frontier models now pose real cybersecurity risks, describing it as having "scrambled the global cybersecurity landscape." Axios places Amodei's proposals as significantly more aggressive than the voluntary 30-day review process in the White House executive order signed in early June 2026. Politico notes it is the most aggressive regulatory framework any major AI CEO has publicly backed. The essay addresses critics who warn strict rules could entrench market leaders - Amodei also flags that both governments and large AI companies must face checks on their power. In an ABC News interview, Amodei said: "We're proposing stronger regulation of the technology, proposing giving the government the ability to, again, in a narrow way, block deployment of unsafe technology."

The essay proposes wage insurance to compensate workers moving to lower-paying jobs, workforce training grants, retention tax incentives for employers, and "universal capital accounts" as a long-term income support vehicle financed through capital gains taxes or levies on AI companies. Amodei also calls for expanded government tracking of AI's labor effects (Axios, Politico).

• •Whether Congress adopts mandatory third-party testing or retains voluntary review, and how "frontier" thresholds are defined - by compute, revenue, or evaluated capabilities (Politico, Axios flag these as open design choices).
• •Which agencies or accredited labs would perform audits and how evidence is shared while protecting proprietary model weights.
• •Whether wage insurance and universal capital account proposals advance in parallel legislation.