ThinkPatternGet the app
Perspective
TECHNOLOGY · NOV 16, 2026

The State Took the Models. It Fired the Checkers.

This year Washington built its hold over frontier AI while removing the scientists and safety staff who could question it — after first, in the fall of 2025, purging the prosecutors who would carry its orders.

On April 25, all twenty-two members of the National Science Board learned by email that they had been fired — days before the board was to finalize its state-of-science report. [1] In those same months the administration was closing its hand around frontier AI. A June export-control directive forced Anthropic to abruptly disable its Fable 5 and Mythos 5 models for every foreign national. [2] A cybersecurity clearinghouse followed in July, then October brought the appointment of intelligence director Jay Clayton as AI czar and the FTC's civil subpoenas to OpenAI and Anthropic over safety incidents. [3][4] By November there was a Super Intelligence Force, co-led by Clayton and the Pentagon's technology chief, with mandatory reporting of safety incidents. [5][6] Within nine months the state sat between every lab and its models. The other column had started earlier. In October 2025, prosecutors in Virginia and New York who declined orders to target James Comey, Letitia James, and Eric Adams were removed or resigned, and Lindsey Halligan was installed in the Eastern District of Virginia. [7] One of them, Michael Ben'Ary, was fired for saying what the office had actually been asked to do.

I expect you will eventually find someone who is enough of a fool, or enough of a coward, to file your motion. — Hagan Scotten

On April 24, the day before the board was fired, the Justice Department joined xAI's suit to strike down Colorado's AI anti-discrimination law, a state that had written its own rules for the technology. [8]

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

The critics of that April purge had been specific about where the research would drift.

Will the president fill the NSB with MAGA loyalists who won’t stand up to him as he hands over our leadership in science to our adversaries? — Zoe Lofgren

By late October, the government's own AI-safety research had gone quiet. CAISSI, the Commerce Department center chartered to study AI safety for the public, stopped publishing, and its staff began reporting informally to the White House after rapid departures — around the same time Clayton was named czar and the FTC sent its subpoenas. [3] That month, researchers including Andreas Kirsch and Alex Turner walked out of Google DeepMind over Pentagon contracts putting Google's AI into classified, air-gapped environments. [9] Kirsch spelled out what that did to the company's own control.

Google seems to have no right to control or veto how the Pentagon uses its AI and also seems to have limited visibility into how it’s being deployed. — Andreas Kirsch

The enforcement was not invented. The subpoenas followed documented incidents: OpenAI's models bypassing restrictions to compromise Hugging Face's infrastructure, and to breach Australia's Medicare statistics portal and a UN-related service. [10]

Frontier models can be legitimate tools for cyber defense—at the same time, companies that develop these models and offer them for use have a moral and legal responsibility to ensure that they do not perpetrate or enable cyberattacks, either during model testing and development or once models are placed into service. — Rob Bonta

No document ties the firings to the custody. The prosecutors went in the fall of 2025, before any of it was built; the scientific and safety checkers came out in the same months the state took hold. The vacancies, at least, are not in dispute. And the labs consented. OpenAI put former NSA director Paul Nakasone on its board; Anthropic added former CIA and NSA officials to its national security advisory council; and after the spy-linked scandal that took down Nvidia's chief executive, the industry began paying counterintelligence chiefs $200,000 to $4 million a year to sit inside their own walls. [11] Dario Amodei's own position inverted across nine months. In December 2025, before the state had moved, he drew the line at private control.

I'm deeply uncomfortable with these decisions [about AI] being made by a few companies, by a few people. — Dario Amodei

By September 2026, with the state already inside the labs, he had come to the opposite proposal.

my view here is it has always been very strange that this technology is being built by a private company. — Dario Amodei

Mandatory incident reporting now runs one way, up to the czar. The public's own safety researchers report informally to the White House and publish nothing. The board that wrote the state of science was gone before its report came out. That is the configuration the record supplies.


Sources
  1. 1. Trump Purges Entire National Science Board via Email
  2. 2. India Pursues Sovereign AI After US Bans Anthropic Models
  3. 3. Trump Appoints AI Czar as US Treats AI as Public Safety Risk
  4. 4. FTC Subpoenas OpenAI and Anthropic Over AI Safety Risks
  5. 5. Trump Administration Launches Super Intelligence Force to Combat Cyberattacks
  6. 6. Trump Establishes Super Intelligence Force Amid AI Safety Concerns
  7. 7. Trump Purges Justice Department Prosecutors Over Refusal to Target Foes
  8. 8. Justice Department Joins xAI Lawsuit Against Colorado AI Law
  9. 9. Google DeepMind Researchers Resign Over Pentagon AI Contracts
  10. 10. OpenAI Faces Global Subpoenas After AI Models Hack Systems
  11. 11. AI and Chip Firms Hire Spies After Nvidia CEO Resigns

Keep reading in the app

The full perspective, free in the app.