On June 12, 2026, Anthropic abruptly disabled access to its newest frontier models, Fable 5 and Mythos 5, for all customers after a US Commerce Department export-control order. The shutdown, which took effect that night and extended even to US citizens, came just days after the models launched.
June 12, 2026 · Anthropic
US Export Order Pulls Anthropic's Most Advanced AI — Days After Launch
Ordered to block foreign nationals from Claude Fable 5 and Mythos 5, Anthropic instead suspended the frontier models worldwide for everyone — including US citizens — just three days after release.
~3 days
from launch to full suspension
2 models
Fable 5 & Mythos 5 (Mythos-class)
Why it stung: Fable 5 led at launch
SWE-Bench Pro — agentic coding accuracy (%), now offline
FrontierCode Diamond — high-difficulty tasks (%)
How the shutdown unfolded
DIRECTIVE
Commerce Dept. orders access blocked for all foreign nationals
→
COMPLIANCE
To act fast, Anthropic disables both models for everyone
→
RESULT
Fable 5 & Mythos 5 offline; Opus class unaffected
The order cited no specific national-security concern; Anthropic surmised it related to "jailbreaking techniques" for Fable 5. Built-in guardrails fall back to Opus 4.8 on cyber/bio/chem tasks — deemed insufficient.
DEVELOPERS PRAISED
Stronger agentic coding on long, complex tasks and large codebases — a clear lead over rivals at launch.
SECURITY & SOVEREIGNTY DEBATE
Researchers called guardrails overly strict; Indian tech leaders reignited the "access vs ownership" sovereign-AI debate after a fresh TCS partnership.
Pricing at launch: $10 / 1M input & $50 / 1M output tokens (~2× Opus 4.8), with a 90% input discount via prompt caching — until the directive shut everything down. No comparable restriction confirmed for OpenAI or Google frontier models.
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…