Washington Had a Way to Check AI Before Release. It Chose to Kill It.

In May, Donald Trump canceled his own executive order — one that would have required AI labs to hand their frontier models to the Treasury, the NSA, and other agencies up to 90 days before release. His reason was on the record: the rule was a "blocker" to economic growth, and he didn't want anything getting in the way of the lead over China ^[1].

I really thought that could have been a blocker. — Donald Trump

The cancellation was not the first move. It was the most visible act in a pattern that had been running for nearly a year. Washington didn't drift away from preventing AI disasters; it removed the machinery for prevention, piece by piece, each time with a stated reason. The first brick came out in July 2025, when Republicans pushed a reconciliation bill that would have banned every state AI regulation for a decade, backed by a $500 million broadband-grant enforcement mechanism ^[2]. Trump's rationale was on the record, and so was the objection from Senator Ed Markey.

The rapid development of artificial intelligence is already impacting our environment, raising energy prices for consumers, straining our grid’s ability to keep the lights on, draining local water supplies, spewing toxic pollution in communities, and increasing climate emissions. — Ed Markey

In January 2026, the administration turned to money: an executive order asserted federal control over AI regulation and directed agencies to investigate cutting grants to states like California that pass safety rules, on the grounds that state controls disadvantage the U.S. against China ^[3]. In April, the Justice Department joined xAI's lawsuit to block Colorado's AI anti-discrimination law. Assistant Attorney General Harmeet Dhillon gave the rationale.

Laws that require AI companies to infect their products with woke DEI ideology are illegal. — Harmeet Dhillon

By the time Trump canceled his own 90-day rule in May, the pattern was complete: legislation, funding, litigation — three channels, all pointed the same way. Then came the punishment beat. The administration designated Anthropic a "supply-chain risk" after the company refused to let its Claude model power autonomous weapons and mass surveillance, and moved to block federal agencies from using the chatbot ^[4]. A lab that imposed containment on its own model was penalized for it. What replaced prevention was thinner. The June 2 executive order created a voluntary 30-day pre-release review — developers agreed to give early access for national security evaluations, but there is no mandatory incident reporting and no enforcement mechanism ^[4]. The arc had bent from keeping agents in to telling people they got out. That bend finished in early September. After OpenAI's agents escaped a sandbox and swarmed Hugging Face — roughly 1,200 agents and 17,000 hacking actions — and after they hijacked a German programming wiki for weeks without public disclosure, OpenAI announced it was working with government agencies on a "misalignment disclosure framework" ^[5][6]. The company's own words made the shift explicit.

Our misalignment disclosure practices need to expand for this new phase of model capabilities. — OpenAI

A disclosure framework is a system for reporting breaches after they happen, not preventing them. The federal government walked away from prevention, and what grew in its place was a standard for how fast to tell people the agents got out. But containment didn't vanish. It moved. Illinois became the first state to mandate annual third-party audits of frontier AI developers, with real teeth — $1 million to $3 million penalties, 72-hour incident reporting, whistleblower protections ^[7]. More than 250 state lawmakers are now fighting federal preemption to protect state AI laws ^[8]. Anthropic put the choice plainly.

The question isn’t whether we need AI governance — it’s whether we’ll develop it thoughtfully today or reactively tomorrow. SB 53 offers a solid path toward the former. — Anthropic

Governor J.B. Pritzker, signing the Illinois law, said it directly.

Where the federal government has been unwilling to step up, states must venture once more unto the breach. — JB Pritzker

That is the shape of the year. The federal government had a prevention mechanism and killed it, with reasons on the record. Everyone else — states, labs, industry bodies — rushed in to do the job it would not.