Trump’s AI Force: Follow the Money, the Power and the Rules

Updated September 21, 2026. President Donald Trump has announced that he is forming an “AI Force” and will name an AI “czar.” The announcement is confirmed. What the AI Force will legally be, where it will sit inside government, what authority it will have, and how it will be funded are still not publicly defined.

What Trump actually announced — and what remains unknown

Trump’s September 19 Truth Social post is the primary source. In it, he said he was forming the AI Force, compared it with the Space Force, said he would soon name an AI czar, argued that the government should not hinder AI growth, and said harmful conduct could be addressed through the existing criminal and civil justice system. He also framed AI as a major economic and geopolitical competition with China and claimed it could eventually account for as much as 25% of U.S. GDP. That 25% figure is Trump’s projection, not an established economic forecast. Primary source: American Presidency Project archive of Trump’s September 19 posts.

Reuters independently confirmed the announcement and reported that the White House had not yet provided implementation details. Axios reported on September 21 that the announcement surprised some advisers and triggered a scramble over what the new entity would actually become. Reuters · Axios.

What already exists before the AI Force

The administration is not starting from zero. In December 2025, Trump signed an executive order directing the attorney general to establish an AI Litigation Task Force. Its stated responsibility is to challenge state AI laws the administration considers inconsistent with a minimally burdensome national AI policy, including laws it believes are unconstitutional, preempted by federal law or otherwise unlawful. The order also directs federal officials to develop recommendations for a national framework while stating that children should be protected, copyrights respected and communities safeguarded. White House executive order.

In June 2026, Trump signed a National Security Presidential Memorandum directing faster AI adoption across military and intelligence functions. The White House says fielded systems are to be robust, steerable and controllable, with clear accountability under the constitutional chain of command. White House fact sheet.

These are documented policies. They should not be confused with powers that the newly announced AI Force itself has not yet been shown to possess.

Why the Space Force comparison matters — and where it stops

The analogy is politically powerful, but it does not establish legal equivalence. The U.S. Space Force became a military service after Congress enacted legislation and the president signed it into law. Trump can create or reorganize some executive-branch functions using existing presidential authority, but a new institution requiring separate statutory powers, major appropriations or military-service status would raise additional congressional questions.

For now, the evidence supports a narrower statement: Trump has announced an AI Force. Its legal form has not yet been publicly established.

Follow the money: four different money trails

“Follow the money” is useful only if the money is separated by type. Campaign spending, nonprofit policy advocacy, corporate infrastructure investment and compliance costs are not the same thing.

1. Political money backing faster deployment

Axios reported in January that the pro-AI super PAC Leading the Future had raised more than $125 million, including support from OpenAI president Greg Brockman, Andreessen Horowitz, 8VC founder Joe Lonsdale, Ron Conway and Perplexity. The group said it would support federal candidates favoring a national AI framework and oppose a patchwork of state regulation. By April, Axios reported that the broader network had raised more than $140 million. Axios, January 30 · Axios, April 16.

A separate organization, Innovation Council Action, was reported by Axios in March to be preparing a campaign of more than $100 million to advance Trump-aligned AI priorities, including infrastructure expansion and lighter federal regulation. It is a separate organization from Leading the Future. Axios.

2. Safety-policy money

Anthropic says it has contributed a total of $40 million to Public First Action. The important legal distinction is that Anthropic says both donations were restricted to Public First Action’s public-education and policy work and cannot be used to influence the election of any candidate. Reuters reports that Public First Action is part of a broader political network that also includes separate PAC activity supporting candidates aligned with stronger AI safeguards. Those are connected political ecosystems, but they are not the same money flow. Anthropic statement · Reuters.

3. Corporate infrastructure money

Federal AI policy affects projects involving chips, data centers, electricity, transmission, land and financing. The administration has highlighted private investment commitments running into hundreds of billions of dollars from major technology companies. Those commitments are not funding for the AI Force itself. They do, however, create a large economic constituency affected by federal rules on permitting, energy, taxes, exports, procurement and regulation.

4. Compliance and competitive costs

Safety rules can reduce risk, but they also create compliance costs. Large frontier labs can generally absorb testing, reporting, legal and licensing expenses more easily than smaller companies. That does not show that safety rules are improper; it does mean policymakers should examine whether particular rules reduce risk, entrench incumbents, or both.

FTC Chairman Andrew Ferguson publicly raised that concern in September. Reuters reported that he said—speaking in his personal capacity—that people should be suspicious when AI companies simultaneously seek new regulation and antitrust exemptions. His argument is one side of the regulatory-capture debate, not proof that any company’s safety position is financially motivated. Reuters.

Who pays for the physical AI buildout?

The infrastructure fight is becoming as important as the model-safety fight. Data centers can require substantial new generation and grid investment. The Trump administration has responded in part through its Ratepayer Protection Pledge. The White House says companies including Amazon, Google, Meta, Microsoft, OpenAI, Oracle and xAI agreed to supply or purchase the power required for their data centers and cover associated delivery-infrastructure costs rather than passing those costs to existing ratepayers. White House fact sheet.

That is the administration’s stated mechanism. Whether it fully prevents cost shifting in every market is a separate empirical question because utilities, state regulators and individual project agreements still matter. Axios reported this month that policymakers and energy leaders are still debating who ultimately pays for new grid capacity created by the AI data-center boom. Axios.

The case for stronger mandatory safeguards

This case does not require assuming Trump is acting in bad faith. It rests on several policy arguments:

Prevention can matter more than liability after a failure. In military, intelligence, cyber and critical-infrastructure systems, critics argue that existing civil or criminal law may punish wrongdoing without necessarily preventing a high-impact AI failure.

Independent testing can expose problems companies miss. A bipartisan Senate proposal reported by Reuters would allow the Commerce Department to demand evidence of safety precautions, use government auditors to test advanced models and potentially seek court intervention against systems considered dangerously unsafe. The proposal remains under discussion; it is not law. Reuters.

Federal preemption can remove state protections as well as reduce duplication. A single national framework may simplify compliance, but critics argue that preemption becomes risky if federal safeguards are weaker than protections already adopted by states.

Political concern crosses ideological lines. Reuters reported that Senator Bernie Sanders and longtime Trump ally Steve Bannon both called for stronger AI oversight in September, despite arriving there from very different political perspectives. That does not mean they agree on the solution; it shows that concern about AI cannot be reduced neatly to one party or faction. Reuters.

The case for faster deployment and a national framework

This argument also has substantive foundations:

A state-by-state patchwork can create real compliance costs. Companies operating nationally may face conflicting obligations. Smaller companies can be especially vulnerable because they have fewer lawyers and compliance staff.

Regulation can unintentionally protect incumbents. Rules that require expensive evaluations, licensing or reporting may improve safety while also raising the cost of entry. Those effects have to be measured rather than assumed.

The U.S.-China technology competition is real. Advanced chips, compute, military applications, scientific capability and AI infrastructure are now strategic assets. Trump’s own September 19 statement explicitly framed maintaining U.S. leadership over China as a central reason not to slow domestic AI development. That is the administration’s strategic argument; whether particular regulations would actually weaken U.S. competitiveness has to be assessed rule by rule.

The administration is not rejecting every safety control. Its June national-security directive requires controllability and accountability for government-deployed AI systems, and the United States has now proposed an AI incident-notification mechanism with China.

Congress is developing a third lane

The congressional debate is not simply “Trump versus Democrats.” Reuters reported on September 14 that Senate Majority Leader John Thune, Senate Commerce Committee Chairman Ted Cruz and Democratic Senator Amy Klobuchar were among the figures involved in discussions over a federal proposal addressing catastrophic AI risk. The proposal could require leading developers to demonstrate safety measures and permit government testing. It also raises the same contested question of whether federal law should override some state AI rules. Reuters.

That is why the eventual statutory language matters more than the political label attached to it. Watch specifically for provisions dealing with independent testing, incident reporting, whistleblower protection, state preemption, liability, open-model restrictions and national-security exemptions.

China: competition and safety cooperation can exist at the same time

The China question is often presented as a binary choice: slow down and lose, or accelerate and accept the risk. Current diplomacy suggests the picture may be more complicated.

Reuters reported on September 20 that Treasury Secretary Scott Bessent proposed a U.S.-China AI safety notification mechanism focused on significant incidents and national-security risks. On September 21, Bessent said the two countries had agreed to continue a formal AI dialogue and meet again in roughly two months, including work on an “incident line.” Reuters, September 20 · Reuters, September 21.

That does not resolve the competition problem. It does show that strategic competition and some forms of safety cooperation are not mutually exclusive.

The real fault lines do not map neatly onto left and right

• Acceleration vs. precaution
• Federal uniformity vs. state authority
• Open models vs. closed models
• Incumbent labs vs. smaller competitors
• Private investment vs. public infrastructure costs
• National security vs. commercial openness
• Voluntary controls vs. enforceable safeguards

Those conflicts can divide companies within the same industry, politicians within the same party and people who agree that AI is both economically important and capable of causing serious harm.

What would make the AI Force concrete?

The next evidence to watch is not another slogan. It is documentation:

• The identity and mandate of the AI czar.
• An executive order, presidential memorandum, charter or legislative text establishing the AI Force.
• Its agency home and reporting chain.
• Its budget and staffing.
• Whether Congress is asked to grant new statutory authority.
• Whether it overlaps with or absorbs the existing AI Litigation Task Force.
• Whether it receives regulatory, procurement, enforcement or national-security powers.
• Ethics, disclosure and conflict-of-interest rules for its leadership.
• Whether federal legislation preempts state AI law.
• Whether model testing and incident reporting become mandatory.

What creators should watch

There is no new “AI Force rule” that ordinary creators have to comply with today. The creator impact will depend on what policy follows the announcement.

For creators, the most relevant questions are whether future federal policy changes:

• access to open-weight or high-capability models;
• synthetic voice and likeness requirements;
• provenance or disclosure obligations for AI-generated media;
• training-data and copyright rules;
• platform and developer liability;
• access to advanced models and APIs;
• tool pricing affected by compute and energy policy; or
• documentation requirements for commercial AI-assisted work.

None of those changes should be assumed from the announcement alone. They are the practical areas worth watching as actual orders, legislation and agency rules appear.

Continue the investigation

• AI Is Not the Wizard: Why AI’s Leaders Are Failing the Human Test
• AI Made It Possible, Part II: The Fight Was Never About the Machine
• AI Made It Possible: If AI Is Going to Kill Us, Show Me the Evidence
• Responsible AI Starts With the Controls You Put Around Your Own Work
• The Trump–Natalie Harp Image Is the AI Lesson We’re Missing

Sources and evidence trail

Primary government and presidential sources

• American Presidency Project — archived September 19 Trump posts
• White House — National Policy Framework for Artificial Intelligence executive order
• White House — AI in the National Security Enterprise
• White House — Ratepayer Protection Pledge

Company and organization disclosures

• Anthropic — Public First Action donation terms

Independent reporting

• Reuters — AI Force announcement
• Axios — internal scramble following announcement
• Axios — Leading the Future financing
• Axios — wider AI influence network
• Axios — Innovation Council Action
• Reuters — Anthropic and Public First Action
• Reuters — FTC chair on antitrust exemptions and AI regulation
• Reuters — Senate safeguards proposal
• Reuters — Sanders and Bannon call for stronger oversight
• Reuters — U.S.-China AI incident-notification proposal
• Reuters — continuing U.S.-China AI safety dialogue
• Axios — dispute over data-center grid costs

Developing story: this page will distinguish new announcements from enacted policy as additional executive orders, legislation, appointments or agency documents become available.