Morning Edition · Tuesday, June 23, 2026
Washington Orders Anthropic to Cut Foreign Access to Its Two Strongest Models
An export-control directive citing a possible jailbreak forced Anthropic to disable Fable 5 and Mythos 5 for all foreign nationals, extending controls from chips to model access itself.

The US government, invoking national-security authorities, directed Anthropic to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States and including the company's own foreign-national employees. Anthropic says it received the order on June 12 and disabled the two models for all customers to comply. Its other models are unaffected.
Commerce Secretary Howard Lutnick notified Chief Executive Dario Amodei by letter without detailing the specific concern. Anthropic's understanding is that the government believes someone found a way to jailbreak Fable 5, meaning a method to bypass its safety restrictions. The company says it disagrees that a narrow potential jailbreak justifies recalling a commercial model used by hundreds of millions, arguing that such a standard, applied across the industry, would halt new frontier-model deployments entirely.
The mechanism is the notable part. Export controls have governed chips and manufacturing tools for years. Applying them to access to a deployed model treats the model weights and serving endpoint as a controlled technology. That expansion of the category affects how every US lab handles non-citizen users and staff. The government's evidence has not been made public, and Anthropic's account is the only source for the jailbreak rationale, so the underlying security claim remains a claim rather than an independently verified fact.
- If true, who benefits
Non-US and Chinese model providers, who can now tell buyers they offer supply Washington cannot revoke, and any party seeking to slow US labs' access to non-citizen researchers and customers.
- The nuance
The directive and Anthropic's compliance are independently confirmed, but the jailbreak rationale comes only from Anthropic's reading of an unspecific letter, and the government has published no technical basis, so its actual motive is undisclosed.
An open-source-intelligence read of how likely this story is true with its real nuance, not a judgment of any outlet. It assesses the claim, weighing independent and adversarial reporting. How we label confidence.
What this means
If a single agency letter can force a frontier model offline for foreign nationals overnight, model access is now a strategic export good subject to the same controls as advanced chipmaking equipment. That raises the operational cost of running global products, complicates the hiring of non-citizen researchers at US labs, and gives non-US providers a reason to tell buyers they offer supply that Washington cannot revoke. Who benefits depends on how the precedent is applied, but the clear loser is the assumption that a deployed model stays available.
What to watch
- Whether the government publishes any technical basis for the directive, which would distinguish a genuine capability risk from a precautionary national-security move and shape how labs contest future orders.
- Whether other US labs receive similar directives, the signal that access-level export control is becoming routine policy rather than a single exception.
Observations to monitor, not financial advice.
Part of a tracked trend
AI Sovereignty and Export Controls on Frontier Models
Over the next 3-6 months, governments increasingly treat frontier AI models as strategic national assets — extending export controls to model access itself and backing domestic 'champion' labs as sovereignty plays.
More from this edition
- SpaceX Leases $6.3 Billion of Nvidia Compute to an Open-Weight Lab, Turning Colossus Into a Merchant Cloud
- OpenAI Ships Full GPT-5.5-Cyber, Pushing Autonomous Vulnerability Hunting Into Defenders' Hands
- Alibaba Open-Sources LOGOS, One Autoregressive Model for Proteins, Molecules and Materials
- Nvidia Now Powers 81 Percent of the TOP500, Tightening Its Grip as Custom Silicon Circles
- Samsung Puts ChatGPT and Codex on 125,000 Desks, Three Years After Banning Them
- Study Finds Transformers' 'Massive Activations' Are Robust, Not a Removable Artifact
- Researchers Show 'Slow-Burn' Prompt Injection Slips Past Per-Event Detectors
- Paper Finds Tree-of-Thought Reasoning Hits a Ceiling You Cannot Buy Past
- Lightweight Prompt Compression Aims to Make On-Device RAG Practical
- Study Says Most Videos Shown to New TikTok Accounts Are Low-Quality AI
- Nvidia Pitches Always-On AI Agents for Running Telecom Networks