Trump says DOJ threatened a major AI executive with jail, but withholds details

Earlier warnings and voluntary commitments

The jail account follows earlier claims of personal control over AI. A report on the interview recounts Trump’s September 14 social-media claim that his administration had stopped AI figures from doing bad or potentially bad things. That post specifically singled out Dario Amodei, chief executive of AI developer Anthropic. The report says Amodei had published an essay warning that development was proceeding unsafely and urging the industry to slow down while governments helped establish safeguards.

The report places those calls in a wider dispute over researchers’ warnings that AI could cause catastrophic harm. It says Trump dismissed calls for a slowdown as a hoax but invited leaders of major AI companies to sign voluntary safety standards at the White House. Trump described the agreement as morally binding; the report says it was not a legal document.

It remains unclear whether the September post referred to the same incident described in the interview. Trump’s earlier criticism of Amodei does not establish that Amodei received the alleged jail threat.

The July 2025 AI Action Plan lists policies on deregulation, infrastructure and security under three pillars. Source: The White House. Source: The White House.

Competition sets the limit on safety rules

Trump’s answer places the federal government in two roles: promoting AI expansion and claiming the power to restrain companies when problems arise.

That tension predates this interview. The White House’s July 23, 2025 announcement of America’s AI Action Plan laid out a federal program to accelerate innovation, build infrastructure and expand international partnerships. It called for removing regulations that hinder AI development and deployment, accelerating permits for data centers and semiconductor factories, and developing the workforce needed to build them. It also called for partnering with industry to export packages of American AI hardware, models, software, applications and standards. These were announced policy actions, not evidence that the promised infrastructure or commercial benefits had been delivered.

In the interview, Trump said the risks were less real than critics suggested and could be managed. He framed AI development as a race with China and argued that falling behind would put the United States at a severe disadvantage. He also claimed government experts were watching the companies, without explaining their methods or powers.

Infrastructure adds another unresolved question. When Douthat challenged him about public anxiety over AI and data centers, Trump claimed new centers were building their own electricity generation and supplying excess power to the grid. His prediction that this would lower electricity costs came without supporting figures in the exchange. The interview provides no basis for measuring that claimed benefit against the power demands of expansion.

There is also political pressure for limits. The report on the interview describes voters passing restrictions on data-center projects in local ballot measures and electing candidates, mostly Democrats, who favored tighter oversight. That opposition gives the debate a constituency beyond AI developers and federal officials.

Yet his answer did leave room for legislation. Asked whether he would work with Democrats on AI safety, Trump said he would. The account of that exchange highlights his accompanying condition: guardrails should not push the United States into a substantially weaker competitive position. He offered no specific legislative proposal in the exchange.

What would turn the anecdote into a precedent

The next policy test is which obligations, if any, Trump would support making binding on AI developers. His willingness to discuss legislation leaves room for a more formal approach, but does not yet establish a safety standard. The distinction between voluntary commitments and legal obligations also leaves an important question unanswered: whether the alleged jail threat concerned an existing law, the safety agreement or something else entirely.

Resolving that question requires the executive’s identity, the conduct at issue, the authority invoked and an account of the resulting changes. Until those details emerge, Trump’s story is evidence of how he presents his oversight of AI companies. It is not enough to assess whether that oversight protects the people exposed to their systems.