Trump Dismantled
Biden’s AI Rulebook
on Day One. Then Built His Own.
On January 20, 2025 — within hours of taking the oath — President Donald Trump revoked Biden’s sweeping AI executive order. On January 23, he signed his own: Executive Order 14179, “Removing Barriers to American Leadership in Artificial Intelligence.” The policy pivot wasn’t subtle. Biden had built a 100-page framework around AI safety, equity, and mandatory red-teaming. Trump replaced it with three words: American AI dominance.
What followed was the most consequential restructuring of U.S. AI policy since the technology emerged as a geopolitical flashpoint. Over the next twelve months, the administration would name a Silicon Valley venture capitalist as its AI czar, announce a $500 billion private infrastructure investment called Stargate, release a 90-point AI Action Plan, convene every major tech CEO at the White House, and attempt to preempt state-level AI regulations in a December 2025 executive order that immediately ran into bipartisan opposition and almost certain legal defeat.
This is the record: what was removed, what was built in its place, who drove it, what the critics said, and where the U.S. stands in the global AI race as of May 2026.
- $500BStargateprivate AI infrastructure commitment announced Jan 21, 2025 — OpenAI, SoftBank, Oracle
- 90+policy actionsacross three pillars in the July 2025 AI Action Plan "Winning the Race"
- Jan 20Day OneBiden EO 14110 revoked within hours of Trump taking the oath of office
Biden’s Executive Order 14110, signed October 30, 2023, was the most comprehensive federal AI directive in U.S. history. At 111 pages, it touched every major federal agency and established what the administration called a framework for “safe, secure, and trustworthy development and use of AI.”
The core provisions Trump’s administration labeled as barriers included: mandatory red-teaming for frontier AI models above defined compute thresholds; mandatory reporting to the federal government by companies developing potentially dangerous models; an emphasis on algorithmic equity requiring AI policies to advance civil rights; and the Biden-era OMB Memoranda M-24-10 and M-24-18, which set AI risk-management requirements for federal agencies. All of it was shelved within hours of the inauguration.
Biden’s EO directed the government to treat AI as a domain requiring safety infrastructure comparable to pharmaceutical approval or financial regulation — voluntary action was insufficient. Trump’s EO inverted the premise: regulation is the barrier, not the safeguard. Where Biden built mandatory checkpoints, Trump issued a 180-day mandate to find and remove them. The legal authority is the same executive pen; the underlying philosophy is opposite.
The rescission was part of a broad first-day sweep. Trump labeled Biden’s AI order “unpopular, inflationary, illegal, and radical” alongside dozens of other orders revoked in the same document. The legal accuracy of those characterizations was contested, but the political velocity was not: EO 14110 was functionally void before the inauguration cake had been cut.
Executive Order 14179, signed January 23, 2025, is three pages — a fraction of Biden’s 111. Its purpose is declarative rather than regulatory. The stated policy: “to sustain and enhance America’s global AI dominance.” The mechanism: get government out of the way.
Three immediate directives came with it. Agencies were told to immediately review all policies, directives, and regulations implemented under EO 14110 and suspend or rescind anything that conflicted with the new policy. The OMB Director was given 60 days to revise Memoranda M-24-10 and M-24-18. And the Special Advisor for AI and Crypto — David Sacks — along with the OSTP director and National Security Advisor, was given 180 days to produce a comprehensive AI Action Plan.
“To maintain this leadership, we must develop AI systems that are free from ideological bias or engineered social agendas.”
Executive Order 14179 — Section 2, Policy — Jan 23, 2025 — whitehouse.gov
The phrase “engineered social agendas” was the tell. It was a direct reference to Biden-era language around AI equity, algorithmic fairness, and DEI considerations in federal AI procurement. Those requirements would subsequently be removed from the NIST AI Risk Management Framework, which the AI Action Plan directed NIST to revise by stripping references to misinformation, DEI, and climate change.
Trump named David Sacks— partner at Craft Ventures and a founding member of PayPal’s “PayPal Mafia” alongside Peter Thiel and Elon Musk — as the White House AI and Crypto Czar. Sacks entered as a Special Government Employee (SGE), which capped his service at 130 days under federal law. He was given access to the White House OSTP staff and co-authorship of the AI Action Plan alongside OSTP Director Michael Kratsios and Secretary of State Marco Rubio.
His 130-day run ended in mid-2025. He then joined the President’s Council of Advisers on Science & Technology (PCAST), where he continued to shape AI and crypto policy without the SGE day-count constraint. His core philosophy was consistent throughout: the private sector, not federal regulators, should determine AI’s trajectory.
“We've got to let the private sector cook.”
David Sacks, White House AI & Crypto Czar — FedScoop interview, 2025
The Biden AI executive order foisted 100 pages of, I would argue, unnecessary burdensome regulation on our AI companies. We're fixing that. The goal is American dominance — not American compliance theater.
Critics at the CSET noted that “some of Washington’s most influential AI policy advocates are also those with the most to gain” from deregulation — a reference to the overlap between Sacks’s venture portfolio and the AI companies that benefited most from lighter oversight. Sacks maintained he was focused on national competitiveness, not personal gain.
The day after the inaugural address, Trump gathered Sam Altman (OpenAI CEO), Masayoshi Son (SoftBank CEO), and Larry Ellison (Oracle Chairman) in the Roosevelt Room. The announcement: Stargate, a joint venture to build AI infrastructure across the United States. Initial committed capital: $100 billion. Target over four years: up to $500 billion. Projected job creation: 100,000 direct U.S. positions.
Ellison told reporters the first data center — a one-million-square-foot facility — was already under construction in Texas. SoftBank holds financial responsibility; OpenAI holds operational responsibility. A fourth partner, MGX (Abu Dhabi), also contributed funding. The announcement drew an immediate challenge from Elon Musk, who posted on X that “they don’t actually have the money” — Altman replied publicly, calling Musk’s claim “wrong, as you surely know.” Trump told reporters Musk’s criticism was about personal animosity with Altman, not the deal’s substance.
This is the largest AI infrastructure project in history. Stargate will create 100,000 jobs right here in America, and it will ensure that we — not China, not anyone else — control the future of artificial intelligence. America is back, and we are going to win.
On July 23, 2025, the White House published “Winning the Race: America’s AI Action Plan” — completing the 180-day mandate of EO 14179. At 25 pages, it outlined more than 90 federal policy actions across three pillars:
Pillar 1 — Accelerate Innovation:Sweeping deregulation. OMB directed to identify, revise, or repeal regulations hindering AI. NIST directed to revise its AI Risk Management Framework, removing DEI, misinformation, and climate-change references. FTC investigations initiated under Biden to be reviewed. Federal agencies to adopt “unbiased AI” procurement standards.
Pillar 2 — Build AI Infrastructure:Expedited permitting for data centers and semiconductor foundries within 180 days. CHIPS Act projects to be freed from “extraneous policy requirements.” DOE and DOD authorized to offer federal land for AI infrastructure. Frontier energy sources — nuclear, enhanced geothermal — explicitly encouraged for AI power supply.
Pillar 3 — Assert Global Leadership:Export America’s “full AI technology stack” to allies as a China alternative. DOC to establish an American AI Exports Program. Allies encouraged to align export controls with U.S. rules to prevent technology leakage. Binding international AI governance agreements deprioritized.
The plan opened with an unambiguous competitive frame: “The United States is in a race to achieve global dominance in artificial intelligence.” China received only two mentions by name in the 25-page document — but the entire architecture was built around the assumption that a Chinese AI lead would be an unacceptable strategic outcome.
The shift is not a matter of degree — it is a reversal of underlying premise. Biden treated AI risk as a market failure requiring government to act as backstop; Trump treats government oversight as the market failure. The following chart captures the eight dimensions where the policy frameworks diverge most sharply.
On December 11, 2025, Trump signed “Ensuring a National Policy Framework for Artificial Intelligence.” The premise: a patchwork of state AI laws was creating regulatory uncertainty that threatened U.S. competitiveness. The solution: federal preemption of state rules deemed “onerous.”
The order established a DOJ AI Litigation Task Force— active January 10, 2026 — empowered to challenge state AI laws in federal court on grounds of unconstitutional burden on interstate commerce, federal preemption, or other legal theories. States with “onerous” AI laws could face loss of certain federal funding. The FTC and FCC were tasked with creating rules that could preempt state regulations.
The order faced immediate resistance from both parties. Florida Gov. Ron DeSantis (R)posted on X: “An executive order doesn’t/can’t preempt state legislative action.” Utah Gov. Spencer Cox (R) called for a version that allowed states to protect children and families. Congress had voted 99-1 against identical preemption language days earlier. Legal scholars at NPR and Axios both noted that executive orders cannot preempt state law without congressional authorization — making the order almost certain to fail in court.
An executive order doesn't/can't preempt state legislative action. Congress could, theoretically, preempt states through legislation. But EOs cannot do this — they only bind the federal executive branch.
“The investment that's happening here — the excitement that I'm seeing from companies that are going to come build here — is going to set us up for a long period of great success.”
Sam Altman, OpenAI CEO — White House AI Summit, September 4, 2025 (White House official release)
“Your Administration supporting our companies instead of fighting with them — it's hugely important.”
Sergey Brin, Google Co-Founder — White House AI Summit, September 4, 2025 (White House official release)
“Some of Washington's most influential AI policy advocates are also those with the most to gain from deregulation — and that conflict deserves scrutiny.”
Center for Security and Emerging Technology (CSET Georgetown) — analysis of Trump AI Action Plan, 2025
The Harvard Kennedy School’s Mossavar-Rahmani Center noted that while the Action Plan’s infrastructure investments are substantial, “dominance framing creates strategic risks” — particularly by incentivizing China to accelerate its own AI development and by alienating European partners who are seeking a middle path between the U.S. and Chinese AI models.
America is the country that started the AI race. And as President of the United States, I am here today to declare that America is going to win it. No other country comes close. We have the best companies, the best engineers, and now we have the plan.
The Trump AI pivot did not happen in a vacuum. By early 2026, three competing governance models were in operation globally. The EU’s AI Act — the world’s first binding AI law, effective August 2024 — established a risk-tiered regulatory regime with mandatory transparency, conformity assessments, and prohibitions on certain AI uses. China deployed a state-directed model emphasizing obedience and security, with mandatory registration of generative AI systems above defined thresholds. The U.S., under Trump, chose a third path: innovation-first, deregulatory, private-sector-led.
The Council on Foreign Relations assessed that 2026 is a pivotal year— when investments made in 2025 produce measurable model capability differences, when allied countries must decide which “AI stack” to adopt for critical infrastructure, and when domestic political pressures in both the U.S. and China will shape whether the technology race accelerates or stabilizes.
The Atlantic Council identified one complexity the Trump administration’s public framing glosses over: the administration has also signaled it will relax export controls on U.S. chips, allowing Nvidia to sell powerful H200s to China — a move that could substantially increase computing power available to Chinese AI firms. The dominance strategy and the export liberalization strategy are in tension, and that tension remains unresolved as of May 2026.
Trump’s AI policy is not a modification of Biden’s — it is its inversion. Biden built safety infrastructure and called it a feature; Trump called it a bug and removed it. What replaced the 111-page Biden order is a 25-page action plan, three executive orders, a $500B private infrastructure deal, and a bet that American companies, freed from “ideological constraints,” will outcompete China before the risks of ungoverned AI compound. Whether that bet pays off — or whether it proves the critics right about what was abandoned — will be visible in model benchmarks, data center maps, and geopolitical alignments long before any administration admits the answer.