Government halts Anthropic’s AI model after safety and regulatory concerns, with officials reviewing high-risk capabilities i

Editorial illustration for Government shuts down Anthropic’s flagship AI after safety warning dispute

Government shuts down Anthropic’s flagship AI after...

Government shuts down Anthropic’s flagship AI after safety warning dispute

By AI Daily Post Edited by Brian Petersen, Editor-in-Chief

June 13, 2026 • 2 min read

The U.S. government moved on Friday to cut off Anthropic’s two flagship models, Claude Fable 5 and Claude Mythos 5, citing national security concerns. Anthropic posted on X that it complied at 5:21 pm ET, but the company argues the order missed the mark.

While the directive forces a worldwide shutdown for all users, it leaves Anthropic’s other models untouched. Mythos, unveiled in early April, is the firm’s most capable system; Anthropic says it can spot security flaws in every major operating system and web browser it examined. Because of that, the model has been limited to a controlled rollout called Project Glasswing, involving roughly 50 vetted partners such as Amazon, Apple, Google, Microsoft and CrowdStrike.

Fable 5, released just three days ago, is essentially Mythos with added guardrails to block high‑risk topics like cybersecurity and biology, making it the most powerful publicly available AI by benchmark standards. The clash raises questions about how export controls intersect with commercial AI releases.

The directive, which Anthropic said it received on Friday at 5:21 pm ET, forces the company to disable both models for all users worldwide — not just the foreign nationals the government’s export control order was nominally aimed at.

— Anthropic, Anthropic’s safety warnings may have just backfired — the government has pulled the plug on its most powerful AI - TechCrunch AI

Why this matters The shutdown of Claude Fable 5 and Claude Mythos 5 puts a stark spotlight on the tension between rapid deployment and governmental oversight. Developers who have built services around Anthropic’s models now face an abrupt loss of capability, and the worldwide disablement means no workaround for foreign users either. Anthropic’s claim that a “narrow potential jailbreak” does not merit a recall raises a question: how narrowly must a risk be defined before regulators intervene?

If the standard applied to the whole industry, as the company warns, future releases could stall before they reach users. Founders must now weigh the cost of integrating third‑party models against the possibility of sudden policy reversals. Researchers lose a testbed that millions were already using, which could slow empirical progress.

We're uncertain whether this move will prompt other firms to pre‑emptively limit model exposure or to double down on safety assurances. In any case, the episode reminds us that technical promise does not guarantee uninterrupted access.