A Trump administration export ban on Anthropic's frontier AI model "Fable" has been lifted after several weeks, following a security agreement between the two sides. Access has been restored, even as the dispute over how to control AI persists.
July 1, 2026 · U.S. Commerce Department
First-Ever Model-Level Export Controls on a Frontier AI — Then Reversed in Weeks
After briefly restricting exports of Anthropic's Claude Fable 5 and Mythos 5 over cyber-misuse fears, the Commerce Department lifted the controls around June 30. Anthropic began restoring global access the next day.
Jun 9
Models released publicly
~3 wks
Duration of export controls before reversal
1st
Model-level controls on a deployed frontier system
How the episode unfolded
Release → Restrict → Restore
1 · Released
Fable 5 & Mythos 5 launch publicly with state-of-the-art benchmarks.
→
2 · Restricted
A jailbreak demo triggers license rules; Anthropic disables access worldwide.
→
3 · Restored
After talks and added safeguards, controls lifted; access resumes.
Benchmark lead
SWE-Bench Pro: 80.3% — 11 points ahead
1,000,000
token context window
$10 / $50
per M input / output tokens (~2× prior)
up to 90%
input discount via prompt caching
Seen as a win
Improved rejection rate for malicious prompts
Usefully sparked regulatory debate
Heavy users welcomed restored access
Raised concerns
Risks slowing progress and creating backlogs
Erodes global trust; pushes sovereign AI elsewhere
Strict guardrails, slow and expensive to use
The bigger picture
U.S. oversight of frontier AI is shifting from hands-off toward security-first — exposing single-provider risk and the hard problem of gating APIs by nationality.
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…