The Trump administration (Department of Commerce) lifted the export controls it had placed on Anthropic's most advanced AI models, Claude Mythos 5 and Claude Fable 5, on June 30, 2026. Anthropic will begin restoring global access starting July 1.
June 30, 2026 · US Commerce Department · Anthropic
US Lifts Export Restrictions on Anthropic's Frontier Models
National-security curbs on the Mythos and Fable models — imposed only about three weeks earlier — were rescinded after a new government agreement, with access restored starting July 1.
~18 days
From restriction (Jun 12) to full lift (Jun 30)
1M
Token context window on both models
2
Models affected — Fable 5 & Mythos 5
Two models, one base — different safeguards
Claude Fable 5
General use · robust safeguards
Blocks or redirects high-risk biology, chemistry & cyber queries. Broadly available via Claude API, AWS Bedrock, Vertex AI, Microsoft Foundry.
$10 in / $50 out per 1M tokens
Claude Mythos 5
Trusted partners only · some safeguards removed
Enhanced cybersecurity capability, incl. vulnerability & zero-day discovery. Limited access via Project Glasswing and select partners.
Same pricing as Fable 5
Why they were restricted
Cyberattack misuse fears (zero-day discovery)
Advanced biology research capability
Cases where Fable 5 safeguards were bypassed
Concerns raised by major investor Amazon
Terms of the lift
Anthropic to proactively detect & address risks
Cooperate on protocols, standards, future releases
Report malicious activity to the government
Preserve US AI competitiveness vs Asian rivals
Timeline of the reversal
Jun 9
Fable 5 & Mythos 5 released
→
Jun 12
Export directive — access suspended for all users
→
Jun 26
Restrictions partially eased for Mythos
→
Jun 30 – Jul 1
Curbs fully lifted; access restored
Praise from developers
Handled most tasks thrown at it, with state-of-the-art claims in software engineering and scientific research.
Criticism
Safeguards called overly conservative — rejecting basic biology questions (e.g. mitochondria); some argued output degradation was anti-competitive.
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…