Anthropic co-founder and Public Benefit lead Jack Clark and the company's Head of Economics, Peter McCrory, discussed AI safety, economic risk and preparation for recursive self-improvement in a conversation published on June 19, 2026 as an Odd Lots podcast and article .
June 2026 · Anthropic
US Blocks Foreign Access to Anthropic's Most Advanced AI Models
Days after launching its top "Mythos-class" models — Claude Fable 5 and Claude Mythos 5 — a US Commerce Department export directive barred foreign nationals from access, forcing a temporary suspension of both models for all users.
Jun 9, 2026
Launch
Fable 5 & Mythos 5 unveiled in the top tier.
→
Days later
Export Directive
Commerce Dept. bars foreign-national access.
→
Response
Suspension
Both models paused for all users to comply.
~5%
of queries hit by Fable 5 guardrails — cyber, bio & model-distillation prompts reroute to the older Opus 4.8.
Shared foundation, different brakes
Claude Fable 5
Claude Mythos 5
Positioning
Safeguarded public version
Cybersecurity use
Availability
API & Enterprise plan
Via Project Glasswing (limited)
Guardrails
Falls back to Opus 4.8 on harmful queries
Relaxed
The Framing
"A gas pedal but no brakes."
Anthropic leadership points to frontier AI's pace and the case for government-imposed development limits. Officials cite national security as grounds for the restrictions — though no specific cases were disclosed.
Capability
Both models lead on reasoning, long-horizon agentic work, coding and vision — with Mythos 5 cited for world-leading cybersecurity.
Concern
Frequent cyber-query restrictions leave the models cutting-edge but practically limited — and the abrupt suspension disrupts corporate operations.
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…