The 'Hoax' Washington Is Already Regulating

The Pentagon keeps a category called "supply-chain risk," built to flag foreign adversaries — the companies and states you don't want inside your military's technology. In February it landed on an American company: the AI lab Anthropic. Nothing Anthropic's models did triggered it. The company was designated because it refused to drop two commitments its own safety policy imposes — no mass domestic surveillance, no fully autonomous weapons. ^[1][2] A list meant for adversaries had caught a rule-writer. The risk the government certified was the safety rules. The pattern holds with every rival who claims a right to write an AI rule. Anthropic's terms, applied to the Pentagon, became an act of usurpation rather than risk management.

The Leftwing nut jobs at Anthropic have made a DISASTROUS MISTAKE trying to STRONG-ARM the Department of War, and force them to obey their Terms of Service instead of our Constitution. — Donald Trump

A state law against AI discrimination in Colorado isn't safety, it's "woke DEI ideology" that the Justice Department joined a lawsuit to kill ^[3][4] — and the White House ordered agencies to explore restricting federal grants to states that pass their own AI rules. ^[5] International oversight at the United Nations is "a globalist scheme." ^[6] The Pentagon's own secretary set the internal register — "Department of War AI will not be woke," no slowdown for "safety-ism" ^[7][1]. And in a final tell, the president ordered "artificial intelligence" struck from official documents and replaced with "super intelligence," because the word artificial makes the technology sound fake. ^[6] Now hold that language against the paper. In June, the same executive built a vetting regime for frontier models — federal agencies get up to 30 days to review a model before release, with classified benchmarking led by the NSA. ^[8] Then, in September, the federal government restricted the release of Anthropic's Fable 5 and Mythos 5 models, with existential-risk mitigation as the stated purpose. ^[9] It was the same month the president called extinction warnings a "hoax" and "a sick conspiracy going on against AI and data centers." ^[10]

We are leading now over China, and everyone else, and I'm going to keep it that way! — Donald Trump

The boomerang is where it sharpens. Anthropic was blacklisted for keeping two limits — no mass surveillance, no fully autonomous weapons. Trump's own June directives prohibit AI used for unlawful surveillance of Americans and require autonomous-weapons policy to respect the constitutional chain of command. ^[8] The limits the lab was punished for keeping now live in the executive's own instruments. When the Pentagon dropped Claude and handed OpenAI the replacement contract, Sam Altman had to amend the deal to bar domestic surveillance outright — the same red line that got Anthropic blacklisted. ^[1] The security rationale was selective all along: through the six-month phase-out, the military kept Claude on intelligence assessments and target selection during joint strikes on Iran. ^[1] The courts have split on what to call it. A California federal judge, Rita Lin, found the blacklisting was likely retaliation for protected speech.

If the concern is the integrity of the operational chain of command, the Department of War could just stop using Claude. — Rita F. Lin

She added that the Pentagon "could just stop using Claude." ^[2] The DC Circuit went the other way, upholding the designation — while conceding the company had been punished for protected speech. ^[7] The labs run the reverse double game. Anthropic's chief executive told senators the risks are real and that governments have "a unique responsibility and capability to act." ^[11] The same industry asked Washington to act as an insurer of last resort.

Given the magnitude of what I expect A.I.’s economic impact to look like, the government should serve the role of "insurer of last resort." — Sam Altman

And Anthropic — the lab the Pentagon blacklisted as a national-security risk — filed for a $965 billion IPO. ^[12] Both double games are real, but they are not symmetrical. That is testimony and a request. The other is a designation, a restricted release, a funding threat against states, and an amended contract. One side asks for the rules; the other side already writes them — and the word "hoax" is what lets it keep writing them while the rest of the system scrambles to write down the danger it insists does not exist.