Trump Rejects an AI Slowdown as Dario Amodei Pushes Frontier Safety Rules

President Donald Trump framed the debate over advanced AI primarily as a competition with China. In public remarks, he said the United States should remain the industry leader because “whoever wins AI wins.” He later called claims that AI could take over the world or destroy humanity a “HOAX” and described opposition to AI and data centers as a “SICK conspiracy” that benefits China. 1719

Why Trump attacked Dario Amodei

Trump’s response followed Anthropic CEO Dario Amodei’s call to deliberately slow the rate at which the most capable AI systems improve. The two positions start from different priorities: Trump emphasized maintaining U.S. technological leadership, while Amodei argued that model capabilities may be advancing faster than effective safety practices and oversight. 17

Trump also argued that existing criminal and regulatory authority was already substantial and said a strong president was the needed safeguard. His criticism of Amodei therefore combined skepticism of existential-risk warnings with concern that formal restrictions could weaken U.S. firms relative to Chinese competitors. 1921

What “We Must Pace the Frontier” proposed

Amodei did not propose ending AI research. His three-part plan called for slowing unchecked capability gains enough for safety work, testing and governance to catch up.

1. Permanent independent evaluators inside frontier labs. Anthropic said it would give third-party evaluators ongoing, employee-level access to its systems. Their role would include verifying safety practices, reporting incidents and assessing alignment during training. 16
2. Coordination among frontier AI companies. The plan called for leading labs to establish shared safety standards and limits on unchecked development rather than competing solely on ever-faster capability gains.
3. International cooperation. Amodei argued that managing the most serious AI risks would ultimately require cooperation beyond one country or bloc.

The first element was presented as an immediate Anthropic commitment, not merely a proposal for governments or rivals to adopt. 16

How other AI leaders responded

OpenAI CEO Sam Altman said he agreed that the frontier should be paced and that OpenAI would likewise provide access to external evaluators. Elon Musk, who runs xAI, also said he agreed with Amodei.

Microsoft CEO Satya Nadella welcomed deliberate pacing and embedded evaluators, but added an important governance condition: oversight should not be controlled by only a small number of companies. He said any pursuit of superintelligence must help humanity and remain under human control.

Those responses do not amount to a complete shared regulatory regime. They do show support among several industry leaders for more formal, continuing scrutiny of the most advanced systems.

The coordination problem: China and unilateral restraint

Former Anthropic and OpenAI researcher Jacob Coxon said people building these systems were “genuinely frightened” about humanity’s future. He welcomed the idea of slowing development, but said it would need coordination with China. 1

That concern goes to the heart of the debate. A company or country that slows on its own may fear losing ground to competitors that do not. China’s foreign ministry pushed back against Amodei’s call for limits affecting China’s AI capabilities, treating it as a U.S. effort to contain China rather than as neutral safety governance. 4

The result is a trust problem: advocates see reciprocal rules as a way to reduce risks, while opponents may see the same rules as a mechanism to preserve another side’s technological lead.

The criticism that safety can protect incumbents

Investor Michael Burry challenged the slowdown campaign as “self-serving.” His critique was that slower development and heavier compliance requirements could help established AI companies by making it harder for challengers to compete. He also disputed the view that current large language models are on a path to artificial general intelligence.

That critique does not disprove AI-safety concerns. It identifies a practical policy risk: rules designed to make frontier development safer can also entrench the firms best positioned to meet them. A credible framework would need to address both safety and concentration of power.

Why the planned Trump-Xi meeting mattered

Trump and Chinese leader Xi Jinping were expected to discuss AI governance at a planned September 24 meeting. 4 The meeting mattered because the hardest part of any frontier-AI safety strategy is not writing a rule for one lab; it is creating arrangements that rival powers believe are reciprocal and enforceable.

Trump’s China-first framing and Beijing’s objection to limits on its AI development suggested how difficult that task would be. The debate is therefore larger than a simple choice between a pause and acceleration: it is a contest over whether the AI race can be governed without turning safety policy into another front in geopolitical competition. 174