The US wants to let AI grow wildly to "defeat China head-on," Trump's policy makes a major U-turn: flattening AI-related regulations across all domestic states

Trump's massive AI policy pivot! On May 26, he appointed former Attorney General Pam Bondi to the President's Council of Advisors on Science and Technology (PCAST), an advisory body chaired by David Sacks and composed of tech titans including Jensen Huang and Zuckerberg. The goal is singular: let AI grow wild and beat China head-on. (Previous context: Trump moves to back CFTC jurisdiction over prediction markets: proper regulation is crucial, many states are abusing regulations to impose sanctions) (Background: Papal encyclical warns about AI: Anthropic's chief admits to fear) Key Highlights - Trump appoints former Attorney General Bondi to PCAST, seated alongside David Sacks, Jensen Huang, and Zuckerberg - The DOJ has established an AI Litigation Task Force to challenge AI regulatory bills in California, Colorado, and other states through federal courts - The White House is using a two-pronged approach combining $42 billion in subsidies and legal litigation, aiming to dismantle AI guardrails in all 50 states to confront China head-on On May 26, the White House released what appeared to be a routine personnel announcement: former Attorney General Pam Bondi will join the President's Council of Advisors on Science and Technology (PCAST). A person whose legal expertise lies in litigation has been placed inside an AI accelerator. Many media outlets only briefly mentioned this matter, but at the same time, Trump indefinitely postponed signing the "Mandatory Review of Frontier Models Order." We can interpret this as Trump's attitude toward AI development—he is dead set on going head-to-head with China. PCAST is co-chaired by former White House AI and Crypto Czar David Sacks and White House Science Advisor Michael Kratsios. The first batch of 15 members includes NVIDIA's Jensen Huang, Meta's Mark Zuckerberg, Oracle's Larry Ellison, Google co-founder Sergey Brin, a16z's Marc Andreessen, AMD's Lisa Su, and Dell's Michael Dell. You can tell from the list, this is the Western world's AI special forces team. This is not an advisory committee that provides suggestions—this is a roster of people who hold global AI computing power allocation rights. The people sitting at this table control the GPUs, data centers, and every supply chain needed to train frontier models. And now, a different kind of person has been added to this table—a former U.S. Attorney General, who can become the strongest legal enforcer. Let's rewind the timeline. On January 20, 2025, within hours of taking office, Trump revoked Biden's AI Safety Executive Order (EO 14110). This presidential executive order required companies to submit red-team test results of frontier models to the federal government. Three days later, Trump signed a new executive order titled "Removing Barriers to American Leadership in AI." On December 11, 2025, Trump signed the "Ensuring a National Policy Framework for Artificial Intelligence" executive order. It contains a highly challenging provision establishing the "AI Litigation Task Force" within the Department of Justice, which from January 10, 2026 onwards, will be specifically responsible for using federal law to challenge state AI regulations. The legal basis is the "Dormant Commerce Clause," because the 50 U.S. states each have their own AI laws, constituting an unreasonable burden on interstate commerce. The White House was then preparing to use federal courts to push back the burdens imposed by all the states. In March 2026, Crypto Czar Sacks's 130-day Special Government Employee term expired, and he transitioned to PCAST co-chair. That same month, the White House released the seven pillars of the "National AI Policy Framework," the seventh of which is "Preempting State AI Laws." Sacks and Kratsios were authorized to draft a federal AI bill, including preemption clauses. On May 26, 2026, former Attorney General Bondi joined PCAST. The enforcer is in position. State-level AI regulation may soon be flattened under the federal government. The last time the United States advanced a technology strategy at this pace was when it detonated an atomic bomb in the New Mexico desert—the Oppenheimer test. - California's SB 53 requires frontier AI developers to publish transparency reports. - Colorado's SB 24-205 prohibits algorithmic discrimination. - Texas's TRAIGA regulates AI governance responsibilities. What these bills have in common is that they are all within the firing range of the DOJ AI Litigation Task Force. An even more ruthless tactic bypasses the courts: the Department of Commerce has been instructed to tie $42 billion in broadband infrastructure subsidies to whether states repeal their AI regulations. If you don't dismantle the guardrails, you don't get the money. The executive order itself technically lacks legal preemption force—it is neither congressional legislation nor a regulation promulgated under congressional authorization. But that has never been the point. Keep state governments sitting in the defendant's chair in federal court long enough, and the cost of litigation itself becomes a form of punishment. The governors of California, Colorado, and New York have stated they will not back down. This battle will ultimately reach the Supreme Court. China mobilizes AI resources uniformly at the national level, without state-level power decentralization. If the United States lets 50 states each pursue their own regulations, frontier models cannot be deployed nationwide—they can't even cross state lines. For the United States to win this AI arms race, the first thing to do is not build more fabs, but first clear away the roadblocks set up by state governments. Part of this is the competitive logic of models. As China massively pushes "decentralized civilian" open-source models, if the United States requires model providers to undergo review before going live, commercializing, or releasing open-source... it amounts to fighting fast with slow. In AI development, fast isn't necessarily right, but fast is definitely valuable. The characteristic of language models is that wild development is the fastest. Trump wants American AI to fully commit to wild development, letting the speed that Silicon Valley startups excel at win the victory. The opponent is too strong, there's no time for democracy—build the thing first, ethics is a post-war concern. AI is the atomic bomb. Many people will scrutinize the ethical issues of the atomic bombing, but few will be dissatisfied with America's victory back then. FAQ Why did Trump put a former Attorney General into the AI advisory team? During her tenure at the DOJ, Bondi signed off on establishing the "AI Litigation Task Force," specifically responsible for prosecuting state AI regulations. After joining PCAST, she can directly coordinate with Sacks, translating policy recommendations into legal action and sweeping away state AI regulatory barriers. PCAST has thereby transformed from an advisory body into a deregulation command center. Which state laws does the DOJ AI Litigation Task Force target? Targets include California's SB 53 frontier AI transparency law, Colorado's algorithmic discrimination ban (SB 24-205), Texas's TRAIGA, and others. The White House is also using $42 billion in broadband subsidies as leverage, demanding states repeal AI regulations deemed "overly burdensome." California and other states have stated they will not back down.