Donald Trump has named the only oversight mechanism he thinks artificial intelligence requires: himself. "The only thing AI needs for control and to set up 'guardrails' is a strong and smart president with a high IQ," he wrote on Truth Social, adding that his administration has already stopped AI company executives from doing "bad or potentially bad" things. He offered no example of any abuse or violation the government has actually blocked.
The post came after a weekend in which the president treated the AI race as a first-order national security matter. On a visit to Ireland he told reporters that whoever wins the AI race wins everything, and he restated the point on social media.
He then turned on Dario Amodei. The Anthropic CEO has called for slowing AI development in the interest of public safety, a position he took after whistleblowers from the company said the technology could "destroy all of humanity" within a decade. Trump accused Amodei of "playing the perfect little angel" and said the administration holds "enormous criminal and regulatory authority" over companies building AI.
Those two sentences do different work, and only one of them is name-calling. The claim of criminal and regulatory authority over AI developers is not a deregulatory statement — it is an assertion of leverage, made in public, against an industry that depends on federal posture toward chips, data centers and power. The president is not saying there will be no oversight. He is saying oversight will be discretionary and personal, applied by him, and that he already applies it. The safety argument he is rejecting is an argument for rules that would bind whoever holds the office next. His answer is that the office is the rule.
Trump also wrote that a "SICK conspiracy" is operating against AI and data centers, that only China is pleased about it, and that the United States leads China and everyone else and will continue to. He put conspiracy theorists, traitors, betrayers and leakers on notice.
Note who that last group is. The safety concern he is dismissing entered public view through company whistleblowers, and the same post that mocks Amodei for moral posturing warns leakers to watch themselves. Whatever one makes of the claim that AI could destroy humanity within a decade, the mechanism that surfaced it is people inside labs telling outsiders what they see. That mechanism is the one now being described as disloyalty.
The unanswered question sits in the first line of the post. If the administration has already prevented AI executives from doing bad or potentially bad things, there is a record of it: a company, a product, a decision reversed. None was named. An oversight regime whose only evidence is the assurance of the person conducting it is indistinguishable from no oversight regime at all — and it is also indistinguishable from one that can be turned on a single CEO who says the wrong thing in public.