The U.S. lets Anthropic ship Mythos again — to about 100 approved companies
Months after a surprise ban, Anthropic's Mythos returns from the regulatory freezer the same day OpenAI previews GPT-5.6. Not for the public: for a short list of “trusted” U.S. firms and agencies. One commenter sized up the caveat as “an asterisk the size of a Mac truck.”
Per Semafor, the government cleared Anthropic to release its frontier Mythos model to a limited set of U.S. companies — reportedly on the order of 100 — while the general public still can't touch Fable. The timing, alongside OpenAI's launch, struck readers as anything but coincidental.
The thread's most uncomfortable thread: regulators routinely restrict products without an act of Congress — cigarettes, explosives, even firearms — so why not models? The question wasn't whether there's precedent, but whether anyone voted for this one.
A frontier model thaws — for a hundred companies, not for you
The Reuters writeup, tracing back to a Semafor scoop, is short on numbers but long on implication: after an earlier surprise ban, the U.S. government has allowed Anthropic to release its Mythos model to a restricted set of American firms. One reader put the scale plainly: “only 100 companies have exclusive access to frontier AI.” And crucially, as another clarified, this covers Mythos access, not Fable — “The general public is still not trusted to use Fable, apparently,” a third noted.
That it landed the same day OpenAI previewed GPT-5.6 was not lost on anyone. A commenter called the “to some US companies” qualifier “an asterisk the size of a Mac truck.”
“Why not models?” — the precedent argument nobody enjoyed conceding
The legal objection came fast: domestic licensing of who may use software should require legislation. “Imposing a licensing system on models for limiting domestic use should require an act of congress,” one reader argued, “but obviously we're well past that red line.”
The most bracing reply came from Slava Akhmechet, founder of RethinkDB, who declined to defend the policy while dismantling the premise:
“Regulatory agencies limit uses of other products without acts of congress — cigarettes, vapes, drugs, pesticides, chemicals, explosives. Even firearms, despite a constitutional amendment! Why not models? … I am making a narrow argument that there is precedent.”
He added the caveat that the delegation of that power should be explicit. The discomfort in the thread was that the precedent is real, and the explicit vote is not.
The endgame read: keep one lab frozen until the other catches up
The most complete theory of the case came from a commenter who'd watched the sequence unfold:
“Back when the administration hit Mythos/Fable with the surprise ban, I figured this would be the endgame. They'd keep Anthropic tied up until a competitor had a roughly comparable model ready, then gate them the same … When OAI wants to release 5.6 wider to actually start making money, I expect Fable will get approved the same day.”
If that's right, the gating isn't about safety pacing at all — it's about keeping the frontier evenly throttled across a handful of approved vendors. The unresolved question, raised by one trader-minded reader: if frontier models can only be sold to a hundred buyers, what happens to the trillions invested on the assumption they'd be sold to everyone?
Get breakdowns like this every day
The top Hacker News stories, unpacked — by email, Telegram, or RSS.
AWS, Google, OpenAI and a dozen more pledge to “defend open source.” The maintainers aren't sold
A Linux Foundation letter announces Akrites: a coalition to find and fix vulnerabilities in critical open source, now that a machine can find a serious bug in minutes instead of an expert's weeks. The first reply asked the obvious question: who actually writes the patches?