Anthropic CEO: Governments should have the authority to block or restrict the deployment of high-risk new AI models.

```

Anthropic CEO Dario Amodei has publicly called for stricter mandatory regulations on artificial intelligence, advocating that the government should have the power to block AI models deemed to pose safety risks from being deployed. This stance differs significantly from the voluntary compliance route currently pursued by the Trump administration.

On Wednesday, Amodei proposed in a lengthy article that AI models should be required to undergo third-party testing, covering cybersecurity threats, bioweapons, and various risk dimensions.

He wrote that if a model is deemed to possess "unacceptable risks," "the government should have the authority to block or restrict its deployment." This is one of his strongest statements on AI regulation to date.

This position stands in stark contrast to current White House policies. Earlier this month, Trump signed an executive order requiring the government to only obtain voluntary access to AI models, without mandating developers to seek explicit approval before deployment.

Regulatory Proposal: From Transparency to Mandatory Constraints

In his article, Amodei used cars, airplanes, and pharmaceuticals as analogies, arguing that AI, like these technologies, is a powerful tool indispensable to the modern economy. However, if poorly designed or operated, it also has the capacity to cause mass casualties.

"Now is the time to move from transparency towards more serious, binding AI regulation," he wrote. He also warned that global policy mechanisms respond slowly, whereas the risks and opportunities brought by AI will "unexpectedly and rapidly compound."

Just last week, Anthropic called for establishing a mechanism by which governments and AI developers jointly decide when to slow R&D to cope with potential dangers. These two public statements show that the company is making the establishment of a regulatory framework one of its strategic priorities.

Mythos Model: Internal Decisions Triggered by Capability Boundaries

Amodei cited Anthropic's new model Mythos in his article, claiming it "perfectly demonstrates both the astonishing capabilities and potential risks of AI."

Anthropic previously revealed that Mythos can identify and exploit critical software vulnerabilities. Accordingly, the company decided to limit access to the model to a handful of partners. On Tuesday, Anthropic released a separate version stripped of cybersecurity functions for wider public use.

This series of decisions highlights the internal contradictions faced by leading AI developers regarding product capabilities and safety boundaries—while publicly calling for regulatory intervention, companies themselves are already conducting similar "regulation-like" internal screening and restrictions.

Policy Differences: The Mismatch Between Industry Calls and Government Direction

Amodei’s remarks position Anthropic in opposition to the current federal policy direction. Trump's executive order favors voluntary mechanisms over mandatory approval, reflecting the administration’s overall approach to reducing regulatory intervention in the AI sector.

Anthropic has long branded itself as a "responsible AI developer." By now publicly advocating for government to obtain mandatory intervention powers, the company signals its intention to play a more active role in driving the regulatory agenda.

However, given Washington's regulatory environment leaning towards relaxation, it remains highly uncertain whether such calls will translate into substantive legislation or regulatory action.

Risk Warning and DisclaimerThe market involves risks. Investments should be made cautiously. This article does not constitute personal investment advice and does not take into account individual users’ specific investment goals, financial situations, or needs. Users should consider whether any opinions, viewpoints, or conclusions in this article are appropriate for their particular circumstances. Investments made on the basis of this article are at your own risk. ```