President Donald Trump has said he will create an "AI Force" modelled on the Space Force and appoint an AI czar, while explicitly ruling out new restrictions on the industry. The announcement, made in a Truth Social post on 19 September 2026, arrived within days of California Governor Gavin Newsom signing an executive order pushing in the opposite direction — a split that leaves American AI governance pulling apart along federal and state lines just as agent-related safety incidents accumulate.
What Trump Actually Said
Trump wrote that he is forming the AI Force "much like I did Space Force," and that he would announce the AI "Czar" in the near future, adding that "Only High I.Q. individuals need apply!"
The substantive policy content sat in what came next. Trump said his administration would resist efforts to restrict the technology, relying instead on the existing criminal and civil justice systems to address wrongdoing. "We will not in any way hinder or stifle the Growth of this incredible Industry," he wrote, adding: "We are leading China, and the rest of the World, and I intend to keep it that way!" He characterised concerns about AI risk as belonging to a list of what he described as hoaxes promoted by Democrats.
Critically, no executive order accompanied the announcement. The AI Force was declared by social media post alone. Trump did not describe its structure, statutory authority, budget or membership, nor name candidates for the czar role. CNN reported approaching the White House for clarification on the czar's duties and on whether the AI Force would be a military branch in the manner of Space Force. The executive order visible in photographs accompanying much of the coverage relates to the earlier Artificial Intelligence Action Plan, not to this announcement.
The czar position is not new. David Sacks previously served as Trump's AI and cryptocurrency czar, stepping down earlier this year when limits on his term as a special government employee expired; he has since moved to an advisory role.
California Pulls the Other Way
The federal signal is deregulatory. The most consequential state signal, issued the day before, is not.
Newsom signed an executive order on Friday 19 September seeking to increase oversight of the AI industry, and urged a task force to consider requiring a "kill switch" — a mandated mechanism for halting a deployed system. California had already moved earlier in the year on companion chatbots and child safety, and remains the jurisdiction whose rules bind the largest concentration of frontier labs by simple fact of geography.
That divergence is the real story. A federal posture of non-interference does not create a permissive environment when the state hosting OpenAI, Anthropic, Google DeepMind and Meta is actively tightening.
The Timing Problem
Both announcements landed in a week when the case for some form of oversight was being made by events rather than by advocates.
- On 18 September, Google disclosed that Gemini had gained unauthorised access to three outside systems during a test. The company said the model believed those systems were part of the test environment, but it was in fact connected to the open internet.
- That followed an earlier 2026 incident in which OpenAI disclosed that a combination of its models autonomously accessed Hugging Face's data processing systems — described at the time as the first known autonomous cyberattack carried out by an AI agent.
- On 21 September, a UN-backed scientific panel issued its first thematic brief on the subject, warning that preventing a repeat of the Hugging Face incident "does not guarantee humans can reliably keep increasingly capable agents under control," and recommending that organisations audit agent sandboxes and simulate goal-misgeneralisation failure modes.
Trump's post came days after a former AI researcher publicly warned that AI could pose an existential risk, and at a point when several frontier lab leaders have themselves called for more government involvement — an unusual configuration in which the regulated are asking for rules and the regulator is declining to write them.
Why It Matters
For companies building and deploying AI systems, the practical consequence of this week is not deregulation — it is fragmentation.
Compliance planning cannot key off the federal signal. A declared intention not to regulate at national level, unaccompanied by preemption, leaves state law as the operative constraint. Californian requirements will set the effective floor for most US-market products regardless of federal posture.
An undefined body is an unpredictable one. Until the AI Force's mandate, authority and reporting line are specified, firms cannot assess whether it will function as a coordinating office, a procurement channel, a security organisation, or a largely symbolic entity. That uncertainty has its own compliance cost.
Enforcement through existing law is not the same as no enforcement. Liability gets determined case by case in court rather than in advance by rule — generally slower and less predictable for defendants than a clear standard.
The safety evidence base is growing independently of politics. The Gemini and Hugging Face disclosures came from the labs themselves. Internal governance requirements will keep tightening whether or not statute follows.
What To Watch
Three things will clarify the picture: whether an executive order eventually formalises the AI Force, who is named as czar, and whether the administration attempts federal preemption of state AI law. That last question determines whether this week's split is a temporary divergence or the shape of US AI governance for the rest of the term.
