The Fable 5 Mess: Anthropic Finally Got the Government Regulation It Asked For

Anthropic spent years telling anyone who would listen that AI was getting dangerous and the government needed to step in. Well, the government finally did step in last Friday, and now Fable 5 is offline, Anthropic is scrambling, and nobody looks happy about it.

Here’s the short version: The Trump administration imposed export controls on Anthropic’s new Fable 5 model and the underlying Mythos model it’s built on. The restrictions said foreign nationals couldn’t access these models anymore — even if they work for Anthropic inside the United States. Anthropic responded by taking both models offline entirely, claiming they couldn’t reasonably comply with the order otherwise.

That’s where things stood as of Tuesday when this Decoder episode was recorded. Fable 5 is still down. Boot up Claude and you’ll see a message right above the chatbox: “Fable 5 is currently unavailable.”

Hayden Field, The Verge’s senior AI reporter, walked through the timeline on the show. She’s been covering this stuff long enough that she has a rule: it’s always a Friday when things go wrong. This was no exception. The chaos started Friday afternoon, continued through the weekend, and by Monday there was still no resolution.

The irony here is thick enough to cut with a knife. Anthropic has been the loudest voice arguing that AI models are approaching dangerous capability thresholds and that the government needs a regulatory framework. Now they’ve got one, and it turns out they don’t love how it’s playing out.

The bigger question — and the one that’ll outlast whatever happens with Fable — is whether this becomes a genuine safety framework or just another tool for the White House to use against companies that don’t fall in line. The Chinese government is watching this closely, and so is everyone else in the industry.

This whole situation raises a question nobody in AI seems to want to answer directly: who decides when a model is too dangerous? Anthropic wanted the government to have that power. Now the government has demonstrated it’s willing to use it. The result is a model nobody can use, a company in damage control mode, and a regulatory precedent that might have consequences far beyond one model release.

I’ve been watching AI regulation debates for years, and this feels like a turning point. Not because the outcome is clear — it’s not — but because the dynamic has shifted. The companies that asked for regulation are now experiencing what regulation actually feels like when it’s not a hypothetical. It’s messy, it’s political, and it doesn’t care about your carefully crafted safety frameworks.

Whether Fable comes back this week or next month or never, the ripple effects are already spreading. The US government just showed it can shut down a major AI model with an export control order. That’s a new reality for the entire industry, and nobody has a playbook for it yet.