The Trump administration on June 26, 2026 authorized AI developer Anthropic to resume limited distribution of its frontier model "Claude Mythos 5," which it had ordered halted, to trusted partners inside the United States, Semaphore reported.
June 26, 2026 · National Security & AI
US Clears Anthropic's Most Advanced AI — But Only for a Trusted Few
After suspending worldwide access on national-security grounds, the administration is letting Anthropic restore "Claude Mythos 5" — limited to 100+ vetted US institutions, in the first substantial move directly regulating an AI model's weights and access.
100+
trusted US companies & government agencies granted access
1st
substantial rule directly regulating an AI model's weights & access
<1
day for red-team testers to build a working exploit from a CVE
Timeline of a sudden lockdown — and a partial reopening
Two models, unveiled together, took very different paths in days.
JUN 9
Mythos 5 & Fable 5 unveiled
→
JUN 12–13
Export directive bars foreign nationals; worldwide access suspended
→
JUN 26
"Safeguards in place" — Mythos 5 restored to 100+ institutions; Fable 5 still restricted
Mythos 5
Defense-oriented · restricted release
Limited to Project Glasswing participants
Cyber-related safety classifiers removed
State-of-the-art in cybersecurity & biology research
$10 in / $50 out per 1M tokens
Fable 5
General-purpose · still restricted
General release, currently held back
Safety classifiers active; offensive requests routed to fallback
Jailbreak potential cited as a concern
$10 in / $50 out per 1M tokens
Why the government stepped in
Potential misuse in cyberattacks
Risk of leakage to China and Russia
Jailbreak potential of Fable 5
What the reopening signals
Prior government clearance now a de facto requirement for AI developers
OpenAI's latest model placed under approval requirement too
Tiered access: only trusted institutions get cutting-edge performance
The trade-off in debate
Top-tier defensive AI for the few — at the cost of broad access and open innovation.
The suspension disrupted research and operations for customers including Fortune 500 firms, while supporters point to genuine national-security risks. The fight now centers on balancing zero-day discovery capability against the freedom to build and ship AI.
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…