Why the US Government Banned
Anthropic's Most Powerful AI Models

The Timeline

What the Directive Actually Says

The order — reported as coming from the White House, citing national security — prohibits use of the two models by anyone who is not a US national. In practice that restricts the most powerful Claude models to American citizens only. Crucially, it doesn't just block users abroad: it also covers foreign nationals physically inside the US, including non-citizen employees at Anthropic itself.

Anthropic complied immediately and pulled access rather than attempt a partial restriction. The company framed it as voluntary cooperation with the government's request.

The Stated Reason vs. the Suspected One

As of mid-June, the government had not publicly stated why it issued the directive. That vacuum has been filled by two competing explanations.

The jailbreak theory

Anthropic's own assessment is that officials learned of a jailbreak — a method for circumventing the safeguards in Fable that stop its most powerful capabilities from being misused. The specific fear, per the company: that the guardrails could be bypassed to extract information useful for cyberattacks. Remember, Fable and Mythos share the same engine; Anthropic describes Mythos as having "the strongest cybersecurity capabilities of any model in the world." A reliable jailbreak of Fable effectively hands that capability to anyone.

Anthropic's public line was notably candid: "perfect jailbreak resistance is not achievable for any current model provider." In other words, this wasn't a fixable bug so much as an inherent property of frontier models.

The "it was never about the jailbreak" theory

Several reporters pushed back on that framing, arguing the jailbreak was a convenient trigger rather than the real cause. The deeper backstory: a months-long dispute between Anthropic and the Pentagon.

Whether the ban is fundamentally about cyber-risk or about leverage in a procurement fight, the two explanations aren't mutually exclusive — and both point to the same new reality.

Why This Matters Beyond Anthropic

This is the first time a frontier AI model was effectively nationalized-by-restriction within days of a public launch. The implications ripple outward:

• Frontier capability is now a controlled export. The most powerful models may ship to the public and then get clawed back on national-security grounds — the way advanced chips and encryption once were.
• "Available" no longer means "dependable." Teams that built on Fable 5 in its three-day window had to rip it out. Model availability is now a geopolitical variable, not just a pricing-page detail.
• Allies are unhappy. A US-nationals-only rule cuts off allied governments and companies that had started integrating the model, straining the same alliances the policy claims to protect.
• Labs face a new tension. Refusing certain government uses (surveillance, autonomous weapons) can carry direct commercial consequences via "supply chain risk" designations.

What Happens Next

Anthropic continues to serve its other models — Opus 4.8 remains the strongest generally available model and is unaffected. Whether Fable 5 returns, and in what form (US-only? with hardened safeguards? as an enterprise-gated tier?), is unresolved. The export-control precedent, though, is unlikely to be reversed: expect future frontier launches to come with access controls baked in from day one.

For the practical question — "what should I actually use right now?" — see our June 2026 best-models breakdown. And for what made Fable so coveted in the first place, read our Claude Fable 5 benchmark review.