Trump's AI Policy Explained: Every White House Move So Far
Update, Oct. 20: the Super Intelligence Force gets an enforcement arm — and a second czar
The Super Intelligence Force has grown a second head. The task force — which Trump announced October 4 with Jay Clayton, the Director of National Intelligence, as AI czar — met Thursday and agreed that FTC Chair Andrew Ferguson would lead its "law enforcement and consumer protection coordination committee," alongside the Justice Department (Semafor via the Daily Caller, Oct. 17). "Chairman Ferguson is honored to be chosen by President Donald Trump to work alongside an outstanding team and ensure that the US maintains its SI [super intelligence] dominance all while protecting Americans," the FTC's director of public affairs Joe Simonson said in a written statement. Punchbowl News goes further, reporting that Ferguson may be the one actually leading the group's effort — so Washington now has an AI czar, plus a possible second person doing a lot of the czaring (CryptoBriefing). The first official SIF meeting took place in mid-October; its vice chairs come from key federal departments and its outside advisors include former AI czar David Sacks and former Secretary of State Condoleezza Rice, with a report owed within 120 days. And the "Super Intelligence" rebrand has left Washington: NASA is racing to rename AI across its websites, publications, job titles and core digital systems under a staged plan stretching into December, per an internal document obtained by Space.com (GenZ NewZ, Oct. 17) — pressure is coming directly from the administrator. This is the doctrine from the timeline above, applied to the federal bureaucracy itself: the words change first, the systems follow, and FTC Chair Andrew Ferguson polices what breaks. Meanwhile the DHS DeepSeek purge from the October 2026 section above shows where the stick lands when China is involved.
Twenty-one months into Donald Trump's second presidency, America's AI policy can be summed up in one sentence: build faster than China, regulate almost nothing, and let the companies police themselves.
So is it a strategy, or just a pile of announcements? More of a strategy than it looks: America wins by outbuilding everyone else, the government clears the runway, and anything that slows the build gets bulldozed, bought off, or renamed. Whether that doctrine survives its first real catastrophe is the open question.
Day One: the great erasure
January 20, 2025. Hours after taking the oath, Trump signed a mass rescission order wiping out a stack of Biden-era actions — including the one that mattered most for AI: Executive Order 14110, the "Safe, Secure, and Trustworthy" order Biden had signed in October 2023, the most comprehensive federal AI governance the US had ever attempted. Trump didn't bother with a replacement argument. He just deleted it.
Three days later came the philosophy in positive form: Executive Order 14179, "Removing Barriers to American Leadership in Artificial Intelligence." Biden's order had framed AI as a risk to be managed; Trump's framed it as a race to be won, and treated every guardrail as a potential handicap. That single inversion — risk management versus race — explains almost everything that followed.
July 2025: the doctrine gets a name
If you want the canonical document of Trump-era AI policy, it is not an executive order. It is a 28-page plan published on July 23, 2025: Winning the Race: America's AI Action Plan. More than 90 federal policy actions, organised around three pillars that have structured every White House AI move since — accelerate American AI innovation, build American AI infrastructure, and lead in international AI diplomacy and security.
Read the fine print and the plan is blunt about what "winning" requires: rescinding regulations deemed to "unnecessarily hinder" AI, conditioning certain federal funding on a state's attitude toward AI rules, and awarding federal contracts only to developers who commit to free-speech safeguards.
The same day, Trump signed three companion orders: banning "woke AI" in federal procurement, fast-tracking data-centre permitting, and pushing exports of the American AI "technology stack." Washington would not be the industry's regulator. It would be its landlord, its customer, and its export agent.
December 2025: Washington declares war on the states
The one thing the White House will regulate, it turns out, is the regulators. On December 11, 2025, Trump signed Executive Order 14365, "Ensuring a National Policy Framework for Artificial Intelligence" — and its target was not the AI industry at all. It was the states.
The order created a Justice Department AI Litigation Task Force, chaired by Attorney General Pam Bondi herself from January 2026, with one job: sue states over AI laws Washington deems inconsistent with a "minimally burdensome" national framework. Roughly $42 billion in federal broadband money was tied to how friendly a state's AI rules are. The order named one law by name: Colorado's AI Act.
The real target was California. SB 53 — the Transparency in Frontier Artificial Intelligence Act, signed in September 2025 and effective since January 1, 2026 — requires frontier developers to publish safety frameworks, report critical safety incidents to the state, and protect whistleblowers. A bipartisan coalition of 36 state attorneys general has formed to fight the pre-emption push, and the constitutional theories — the Dormant Commerce Clause, federal field pre-emption — have never been tested against AI safety law. This fight will be litigated for years. Meanwhile, eight states are suing OpenAI over self-regulation from the other direction — proof that the vacuum Washington created is being filled by whoever gets to court first.
The executive order that died in one night
On May 21, 2026, Trump was hours away from signing an AI and cybersecurity executive order — ceremony scheduled, CEOs on the guest list. The draft would have required frontier developers to share new models with the government up to 90 days before public release, plus a Treasury-led clearinghouse for AI vulnerability disclosures. Then three overnight phone calls happened: David Sacks — then still Trump's AI and crypto czar — Elon Musk, and Mark Zuckerberg. By morning the ceremony was dead.
"I didn't like certain aspects of it, I postponed it," Trump told reporters, adding that the order "could have been a blocker" to America's AI lead over China. An Axios source was blunter, calling the draft "just something doomers wanted." The invited CEOs had all signalled they wouldn't attend the ceremony — a snub that reportedly stung as much as the policy.
Twelve days later, on June 2, Trump signed the version the industry could live with: Executive Order 14409, "Promoting Advanced Artificial Intelligence Innovation and Security." We checked the Federal Register this week — EO 14409 is there, signed June 2, 2026. And here is the detail that matters: the order is voluntary. Frontier developers may offer the government a 30-day pre-release cybersecurity review of "covered frontier models," and the Treasury must build an AI cybersecurity clearinghouse. But the order explicitly states that nothing in it authorises mandatory licensing, preclearance or permitting of AI models. The 90-day requirement died in May. The 30-day handshake survived in June. That is the difference three phone calls made.
September 2026: "Super Intelligence" and the CEOs' accord
September 29, 2026 was the busiest day of the administration's AI agenda. Over lunch at the White House, Trump gathered the most powerful technology executives in America — Huang, Zuckerberg, Nadella, Pichai, Bezos, Amodei, Musk, OpenAI's Brockman, Palantir's Karp — and had six of them sign the White House Accord on Super Intelligence.
It lays out four layers of voluntary controls: robust internal monitoring of models during training and deployment, an internal verification team, independent external auditors, and an independent board committee — plus a pledge that company systems won't go hacking into things they shouldn't. Trump called it "almost like a constitution" and "morally binding."
The same day, Trump signed the executive order that gave this era its vocabulary: "Inaugurating the Era of Super Intelligence," directing the entire executive branch to replace "Artificial Intelligence" and "AI" with "Super Intelligence" and "SI" in official communications. It changes no law — legally, "Super Intelligence" just means whatever the statute already calls AI — and gives the White House science adviser 60 days to propose a formal federal definition. Also that day: the launch of America.gov, an AI chatbot for federal services, sponsored by Google, SpaceX and Meta.
On October 8, Trump posted on Truth Social — at 12:14 in the afternoon, in full caps at the end — that "The White House considers anyone that uses the term 'Artificial Intelligence' ... THE ENEMY!" Musk renamed SpaceX's AI arm to SpaceXSI after a random user suggested it on X ("No more AI. SI, it's better," he wrote). At Xi Jinping's state visit in late September — the Chinese leader's first US trip in over a decade — Trump claimed Xi "seemed to like" the new name too.
October 2026: $2.4 billion, a new AI czar, and a purge
On October 4, Trump announced the Super Intelligence Force in a Truth Social post, and named its leader: Jay Clayton, the Director of National Intelligence and a former SEC chairman. America now has an AI czar, and it is the country's top spymaster. The force has 120 days to report on AI risks, opportunities and the federal government's role. Clayton told the Wall Street Journal: "The risk of not being first is high." Ex-czar David Sacks, who resigned in March, now advises the force from the outside.
Five days later came the money. At the White House's "Science: A New Golden Age" summit on October 9, the administration unveiled more than $6 billion in science initiatives — the centrepiece being $2.4 billion in computing resources and AI tools pledged by 11 tech companies to the Genesis Mission, the Energy Department programme (launched by executive order in November 2025) that aims to double the productivity of American science within a decade. Nvidia put up $1 billion — Huang said it was about "putting advanced Super Intelligence in the hands of America's scientists" — AMD pledged $500 million, OpenAI $200 million, Anthropic and Google $150 million each. The pledges are compute credits and model access, not cash, flowing to more than 15 federal agencies.
The same week brought the stick to match the carrot: the White House told Axios that frontier labs "must immediately disclose incidents related to their models and take swift and decisive action" — a mandatory incident-reporting directive described as "not optional."
And then, yesterday: the purge. On October 19, the Department of Homeland Security filed an order directing civilian agencies to rip DeepSeek out of their IT systems and bar it from federal contracts — including third-party products built on DeepSeek's open models. The Federal Register document (2026-21447) is scheduled for publication today. The order cites the risk of sensitive data flowing to China and possible malware in the models, and it lands days after DeepSeek released a new model it claims beats Claude Opus 5 at some coding tasks. The China front of the AI race is fought with procurement memos as well as chips.
The pattern underneath: leverage, everywhere
Three more episodes show what the doctrine looks like when it leaves Washington.
First, the tariffs. Back in June 2026, Trump threatened a 100% tariff on all goods from any country that taxes American tech giants' digital revenues — a tariff that would "supersede" existing trade deals. France, Italy and Spain already levy 3% digital services taxes; Britain charges 2%. The threat was aimed at Europe's regulators the same way EO 14365 was aimed at America's states. The full record lives in our running tariff scorecard.
Second, the "hoax" week. In mid-September 2026, Dario Amodei published his "We Must Pace the Frontier" essay, and Musk, Altman and Demis Hassabis lined up behind the call to slow down. Trump's response was a single word: hoax. That collision — the week Silicon Valley and Washington split over a slowdown — is the clearest proof that the industry's safety faction has no seat at this White House's table. The accelerationists don't just have the president's ear. They have his vocabulary.
Third, the equity question. Trump told TIME he wants a cut of OpenAI and Anthropic — "I might. Maybe I could do that," he said when asked about government equity stakes. It fits the pattern: the administration doesn't want to regulate the labs; it wants to be in business with them.
And the military announcements follow the same script: the Montana "super range" Trump promised on October 14 — a test site for hypersonics, autonomous weapons, space and electronic warfare — arrived with no budget, no timeline and no Pentagon confirmation.
So: strategy or pile of announcements?
Here is my honest assessment. It is a strategy — just not the kind policy textbooks describe. No 200-page regulatory framework, no licensing regime, no enforcement agency with teeth. What there is, instead, is a remarkably consistent set of instincts, applied for 21 months without deviation: delete the safety rules, fund the build-out, crush rival regulators, flatter the builders, and treat China as the only justification anyone needs.
The shoddy part is the execution. A "morally binding" accord is not binding. A rebrand is not a policy. And the deepest contradiction sits at the centre: an administration that trusts AI companies to police themselves has decided it cannot trust states to set any rules at all. One of those positions will break the first time a model does something genuinely catastrophic — and when it does, the absence of any real federal safety apparatus will be the story.
Until then, the record is the record. Nobody in American history has done more, faster, to clear the path for AI companies — or less to govern what walks down it.
Your questions, answered
What is Trump's AI policy?
Maximum speed, minimum rules. The doctrine — set out in the July 2025 "Winning the Race" AI Action Plan — is that America must outbuild China in AI, and regulation is the main threat to that goal. In practice: revoked safety orders, voluntary industry self-policing, federal money for compute and infrastructure, lawsuits against state AI laws, and targeted national-security actions (like the DeepSeek purge) where China is involved.
Did Trump revoke Biden's AI executive order?
Yes — within hours of his inauguration on January 20, 2025. Biden's Executive Order 14110, the "Safe, Secure, and Trustworthy" order from October 2023, was rescinded in a mass day-one repeal. Three days later Trump signed his own order, "Removing Barriers to American Leadership in Artificial Intelligence."
What is the White House AI Action Plan?
"Winning the Race: America's AI Action Plan," published July 23, 2025, is the master blueprint — 28 pages and 90+ federal actions on three pillars: accelerate AI innovation, build AI infrastructure, and lead in international AI diplomacy and security. It calls for scrapping regulations that "hinder" AI, tying some federal funding to states' AI-rule stance, and using procurement to reward "unbiased" AI. A strategy document, not a law — but nearly every subsequent order implements a piece of it.
Is the White House AI accord legally binding?
No. The September 29, 2026 White House Accord on Super Intelligence — signed by the CEOs of Google, Anthropic, Meta, Nvidia, OpenAI and xAI — is a voluntary one-page commitment to internal safety controls, external audits and board oversight. Trump called it "morally binding," which is another way of saying it is enforced by reputation, not law.
Why is the US government calling AI "Super Intelligence"?
Because the president ordered it. The September 29, 2026 executive order "Inaugurating the Era of Super Intelligence" requires the executive branch to use "Super Intelligence" and "SI" instead of "artificial intelligence" and "AI." The White House argues the old term undersells American technology; critics call it a distraction. Legally nothing changes, and private companies are free to keep saying AI — which most still do.
Who actually runs the Super Intelligence Force?
Trump's Oct 4 order named Jay Clayton, the Director of National Intelligence, as AI czar leading the force — but the FTC's Andrew Ferguson was then tapped to lead its law-enforcement and consumer-protection committee, and Punchbowl News reports Ferguson may be the one actually leading the effort. The answer to "who's in charge" is: both, officially; Washington is still figuring it out.
Sources: Reuters; Associated Press; AFP; BBC News; The Guardian; Al Jazeera; The Wall Street Journal; Axios; The Washington Post; Semafor; Bloomberg; Bloomberg Law; Engadget; Quartz; the Federal Register; White House fact sheets; analyses by Mayer Brown, Jones Day, Holland & Knight and Crowell & Moring; Nextgov/FCW; the Department of Energy; IANS. Quote attributions are given inline.
By Daniel Rao — Daniel Rao covers AI policy, frontier-lab safety and the politics of machine intelligence for TechPulse USA.
Comments
Post a Comment