The US Commerce Department on June 12, 2026 placed Anthropic's flagship models Claude Fable 5 and Claude Mythos 5 under export controls, barring use by foreign governments, companies and individuals. Anthropic suspended all user access to both models the same day to comply.
June 12 · US Commerce Department × Anthropic
Washington Orders Anthropic to Cut Off Its Top AI Models — and They Go Dark Worldwide
An export-control directive demanded Fable 5 and Mythos 5 be suspended for "any foreign national." Unable to fence off only foreigners, Anthropic abruptly disabled both frontier models for all customers, everywhere.
2
frontier models disabled
5:21pm ET
directive received — no specifics on the security concern
Global
cutoff for every customer — only foreign carve-out impractical
What got pulled
A generational leap, frozen — SWE-Bench Pro (agentic coding)
FrontierCode Diamond 29.3% vs 13.4% / 5.7%
Terminal-Bench 2.1 88.0% vs 82.7%
Humanity's Last Exam 59.0% vs 49.8%
The dispute
Triggered by a narrow jailbreak on Fable 5 with limited cyber implications — which Anthropic calls a "disproportionate" response.
The flaw is minor and already known in other models such as GPT-5.5.
Fable 5 underwent extensive red-teaming before release.
If applied broadly, similar action could halt new frontier deployments industry-wide.
Before the cutoff — the models
Developers rated them state-of-the-art for long-horizon agentic coding, software engineering and complex analysis. Priced at $10 / $50 per M tokens — under half of Mythos Preview.
After the cutoff — the backlash
Shock at the abrupt global shutoff, warnings of "AI nationalism," and concern it is counterproductive for US competitiveness and upcoming AI IPOs.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…