Breaking

US Lifts Export Curbs on Anthropic's Fable 5 and Mythos 5

July 1, 2026 at 13:27 EDT

Policy & Regulation
Foundation Models
Security

On the night of June 30, 2026, the Trump administration lifted the export controls it had imposed on Anthropic's AI models "Claude Fable 5" and "Mythos 5." Anthropic said the same day it had been notified by the US Department of Commerce and announced it would begin restoring access globally starting Wednesday, July 2.

June 30, 2026 · Anthropic

US Lifts Export Controls on Claude Fable 5 & Mythos 5

The Commerce Department reversed restrictions imposed just over two weeks earlier. Anthropic began restoring access from July 1 — Fable 5 for general users, Mythos 5 in stages for trusted partners.

A Hectic 3-Week Rollout

Jun 9

Models first released

→

Jun 12

Export controls imposed; models disabled

→

Jun 30

Controls lifted after safety fixes

→

Jul 1

Access restored in stages

18 days

from controls imposed to lifted

token context (up to 128k output)

Jan 2026

knowledge cutoff

Pricing per Million Tokens

Output costs 5× the input rate.

$10

Input

$50

Output

Claude Fable 5

Safety-hardened, general-use model. Restored for all users, inside and outside the US.

Claude Mythos 5

Less-restricted variant for cyber and biology fields. Offered in stages to trusted partners only.

Praised

State-of-the-art in software engineering, knowledge work, vision and science — strong on long-running, agentic tasks. One early reviewer said it "handled nearly every task."

Criticized

New safeguards trigger frequently and curb capability — called impractical for frontier LLM work, alongside complaints of high pricing and slow processing.

The reversal followed added safety classifiers and talks with Commerce leadership.

The policy backdrop: balancing stronger AI export controls against preserving the competitive edge of US firms, as rivals accept similar user vetting.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.