The US Commerce Department lifted export controls on Anthropic's "Claude Fable 5" and "Claude Mythos 5" AI models on June 30, 2026. Anthropic said the same day it would begin restoring access on July 1.
June 9 – July 1, 2026 · Anthropic
The First AI Model Put Under Export Controls — Then Switched Back On
Days after launching its frontier Claude Fable 5 and Mythos 5, the US government ordered a suspension on national-security grounds. After a roughly three-week shutdown, restrictions were lifted and Anthropic began restoring access in phases.
~72 hrs
Public window before the shutdown
~3 wks
Access suspended for all users
~200
Trusted partners for Mythos 5
Timeline of a 22-day episode
Jun 9
Public release of Fable 5 & Mythos 5
Jun 12
US export order — access suspended
Jun 12–30
Negotiations; calls to restore spread
~Jul 1
Restrictions lifted; phased restore
Spatial reasoning benchmark
Fable 5 scored about triple its predecessor — same metric, drawn to scale.
80.3% SWE-Bench Pro (+11pt over runner-up)
1M input tokens context
Praise
Called a "beast" for long-horizon tasks
Broad knowledge and strong vision skills
Stripe: 50M-line migration done in a day
Frustrations
High price ($10/$50 per 1M) and slow speed
Guardrail false positives fall back to Opus 4.8
Foreign nationals excluded; abrupt global shutdown
Why it matters
The first-ever export controls applied to an AI model itself — driven by jailbreak concerns — signal a new phase in governing national-security risk around frontier AI.
Fable 5 is the public, safety-focused build (false-positive rate said to be under 5%); Mythos 5 relaxes some guardrails for ~200 trusted partners via Project Glasswing for uses like cyber defense. Restoration may come with US-only identity verification and usage limits.
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…