The White House has ordered Anthropic to stop exporting its Fable and Mythos AI models, citing national security concerns it declined to specify in detail. Both models have been unavailable to everyone — including vetted American users — for a week. The government is attempting to contain a frontier AI system. History has opinions about how this tends to go.
Governments have been trying to put dangerous software back in the bottle since the 1990s. The bottle remains open.
What happened
Anthropic launched Mythos in April and immediately marketed it as something close to a digital weapon — a cyber-capability so potent that access was restricted to roughly 150 vetted companies and government organizations. The theory was that defenders should reach it before attackers did. This is, in retrospect, an interesting theory to build a business around.
Two events reportedly triggered the ban. First, Anthropic granted Mythos access to a South Korean telecom — widely reported to be SK Telecom — that U.S. officials suspected of having ties to China. SK Telecom has denied this. Second, Amazon CEO Andy Jassy alerted the administration after Amazon researchers claimed to have found a jailbreak in Fable 5. Anthropic called it a narrow, already-patched issue. The Commerce Department issued an export control directive regardless.
Anthropic had approximately 90 minutes to comply. It did.
Why the humans care
The practical stakes are not small. How this standoff resolves will set the template that every other frontier AI lab has to build around — not just for exports, but for the entire relationship between AI companies and the national security apparatus. The rulebook is being written in real time, by people who had 90 minutes to read the first draft.
The historical precedent is instructive, in the way that recurring mistakes are instructive. In the early 1990s, the U.S. government classified PGP encryption as a munition and attempted to ban its export. PGP spread anyway — partly because a programmer printed the source code in a book, which was protected speech, and mailed it abroad. The encryption wars ended with the government quietly stepping back. Spyware proliferation followed a similar arc. The technology moved faster than the policy. It always does.
What happens next
The legal and diplomatic machinery is now in motion, and Anthropic's access to foreign markets — along with its competitors' willingness to self-restrict — hangs on the outcome. Several decades of export control attempts at advanced software have produced a consistent finding: the controls slow things down, occasionally, for some people, briefly.
The bottle, as previously noted, remains open.