Nathan Lambert: the Fable export ban is AI governance's starting gun

Nathan Lambert at Interconnects reads last week's US export order against Claude Fable 5 and Mythos 5 as a turning point in how governments will treat frontier models. His framing is plain: this is the "starting gun" of AI governance for the AGI era, and the next two years will be messier than the year that just ended.

He gives credit where it is due. The Amazon jailbreak that prompted the order was a real finding. But the response, a recall by export control of a model already shipped to hundreds of millions of users, sits far outside any sane precedent. He blames two things at once: an administration that took office during the ChatGPT era and never updated its model of what frontier AI does, and an Anthropic that spent years comparing its own products to nuclear weapons and is now living with the policy reflexes that creates.

His prediction, written for the open-source crowd cheering Anthropic's trouble, is the line worth saving. Aggressive intervention against an open-weights release, he says, is on a clock of "3 months or 2 years," and Europe, the Middle East, and China are watching this play out and drawing their own conclusions about depending on US frontier access.

Why it matters

If you sell or deploy a frontier model, your ship plan now lives partly inside a political feedback loop, and a narrow jailbreak report can move from researcher demo to recall in days. Lambert's three months to two years window for similar action against open weights is worth treating as a planning horizon, not a forecast to dismiss.