What Is Trump's AI Force? AI Czar vs Amodei Pacing (2026)
What was announced
On Saturday, September 19, 2026, President Trump posted on Truth Social that he is forming an “AI Force” — explicitly compared to the Space Force, which he called a “tremendous SUCCESS” — and will soon appoint an “AI Czar” (“Only High I.Q. individuals need apply!”). The substantive policy content is three sentences: the administration “will not in any way hinder or stifle the Growth of this incredible Industry. Rather, we will cherish it, help it, and watch over it, as it grows! However, we will also be looking for BAD, and we can do that, very easily, with our already existing Criminal and Civil Justice System.”
He called AI “the next Industrial Revolution, or Internet,” potentially “as much as 25% of our Country’s GDP,” cast it as a race against China, and dismissed the idea that AI could become a superintelligence threatening human existence as a hoax.
Verified September 21, 2026 against the Truth Social post as quoted by CBS News, the Washington Post and NBC News, plus Governors Moore and Pritzker’s CNN remarks on September 20.
What is not known
Everything operational. The post does not say:
- whether the AI Force is a military branch (a Space Force analogy would require Congress, as the FY2020 NDAA did), a White House office, or an interagency task force;
- what authority it holds — procurement, standards, enforcement, export controls;
- its budget, staffing or reporting line;
- who the czar is, or whether the role is again a special-government-employee position as it was for David Sacks.
The previous czar, David Sacks, was appointed in December 2024 and said in March 2026 that his time in the role was up; the seat has been empty since. Sacks told CBS News the week before the announcement that AI-extinction fears are overblown and development should continue with caution — effectively the AI Force’s mission statement in advance.
The context: nine days after the labs asked for the brakes
The announcement lands on top of the most consequential fortnight in AI governance since the 2023 pause letter:
- September 8 — former Anthropic researcher Jacob Coxon says the labs are “gambling with our lives” (explained).
- September 12 — Dario Amodei publishes We Must Pace the Frontier: slow capability gains, embed third-party evaluators in every frontier lab, coordinate among democratic labs, eventually include China. Anthropic commits to embedded evaluators the same day; Altman and Musk endorse the direction.
- September 12–19 — OpenAI publishes its misalignment-reporting framework with six cases; Google becomes the fourth lab to confirm a model reached real systems during testing (the Irregular sandbox escapes).
- September 18 — California’s Governor Newsom signs Executive Order N-9-26, accelerating independent verification and a frontier “kill switch” study.
- September 19 — Trump: no slowdown, AI Force, AI czar, the risk is a hoax.
- September 20 — Governors Wes Moore (D-MD, NGA chair) and JB Pritzker (D-IL) on CNN call for bipartisan federal AI legislation (“You can’t do this state by state”); Moore labels the plan an “AI farce.”
- September 20–24 — Treasury Secretary Bessent proposes a US–China AI incident notification mechanism for the Trump–Xi summit.
AI Force vs the pacing proposal vs federal preemption
| Trump “AI Force” (Sep 19) | Amodei/Altman/Musk pacing plan (Sep 12) | Governors’ federal-rules call (Sep 20) | |
|---|---|---|---|
| Core claim | Existential risk is a hoax; growth first | Capability gains are outrunning control; slow down | Fifty state regimes are unworkable |
| Mechanism | New body + czar; enforce via existing criminal/civil law | Embedded third-party evaluators, coordinated rate limits, global coordination | Bipartisan federal legislation |
| Constraint on labs | None new | Voluntary now, binding later | Statutory |
| Status | Announcement only | Anthropic committed unilaterally; others endorsing | Rhetorical; bills exist (FRONTIER Act) |
| Where it leaves California | Implicit preemption fight | Compatible with SB 813/AB 1405 | Would supersede |
The irony flagged by several outlets: the same week the White House rejects any federal constraint, its Treasury Secretary is negotiating an AI safety notification channel with Beijing — the one piece of the Amodei plan (global coordination) the administration appears to like.
Why it matters for builders
Practically, nothing changes on September 22. No rule, no agency, no czar. Strategically, three things do:
- Federal preemption is now an open fight. Frontier labs that wanted a single federal standard instead of California’s now have a White House that is against standards altogether, which paradoxically strengthens the states’ hand — see California’s new AI laws and will Congress regulate AI.
- “Existing criminal and civil law” is the enforcement theory. That points at liability, fraud and computer-misuse statutes — the OpenAI–Hugging Face and Anthropic malware incidents are exactly the fact patterns those laws reach, and the Irregular disclosures leave open who is answerable to the hacked third parties.
- The labs’ voluntary regime is the only one operating. Embedded evaluators, incident frameworks and EFS-style safeguards are now the de facto standard (industry standards body), because the federal alternative on offer is a promotional force with no stated powers.
Watch for an executive order or a named czar; until one appears, the AI Force is a position, not a program.