TL;DR
The Commerce Department's export directive blocks foreign access to Anthropic's newest models after an unnamed company claimed it bypassed Fable 5's safety restrictions.
At 5:21 pm Friday, Anthropic shut off worldwide access to two of its newest models. Claude Fable 5 and Claude Mythos 5 are now unavailable for all customers, following an export control directive that Commerce Secretary Howard Lutnick addressed personally to Anthropic CEO Dario Amodei.
Both models had launched days earlier. Fable 5 was the public product, priced at $10 per million input tokens, designed with guardrails to contain the most dangerous capabilities of its parent architecture. Mythos 5 remained restricted to Project Glasswing, a curated program for security research partners. During the Glasswing preview, Mozilla alone reported patching hundreds of vulnerabilities using Mythos Preview.
According to 9to5Mac, the Commerce Department's order requires a license for any transfer of these models outside the United States and restricts access for foreign nationals inside the country. The Trump administration had attempted to prevent the release before launch and failed. The export control mechanism is the lever it found that did not require prior consent.
What the jailbreak claim involves
An unnamed company told officials it had found a method to bypass Fable 5's safety restrictions, raising administration concerns about national security implications. Anthropic responded in a blog post arguing the government's reaction is disproportionate to the actual risk. The company has not confirmed what specific capability the exploit exposed or how reproducible it was.
Forbes identified the central tension: Anthropic had itself warned in April that Mythos was "too dangerous" for broad release, citing potential damage to economies, public safety, and national security. Fable was presented as the responsible alternative. That safety narrative, authored by Anthropic, handed policymakers the vocabulary to justify this intervention.
Policy logic is not arbitrary when a company's own documentation describes severe national-security risks from artificial intelligence. If you build your corporate identity around existential-level safety warnings, eventually a government will treat those warnings as actionable.
Anthropic is complying with the directive, per Bloomberg, while signaling disagreement. That tension carries commercial weight. Fable 5 was already live on Azure and Amazon Bedrock, with customers who had integrated it into long-context reasoning and security research workflows.
Spillover into open-source
Crypto Briefing covered analyst commentary pointing to a predictable second-order effect: restricting access to US frontier models pushes organizations toward open-source alternatives, several of which originate in Chinese labs and carry separate governance questions. Export controls on proprietary systems can inadvertently strengthen the competitive position of those alternatives.
For practitioners who had planned deployments around Fable 5, no migration path has been announced. The 1M token context window and competitive pricing had made it attractive for exactly the use cases now off-limits. Price Per Token flagged the developer community's reaction, with commentary framing this as the first instance of a US government body exercising real-time distribution control over a frontier model after release.
A harder jurisdictional question follows. Export control frameworks were built around hardware, not API-accessible artificial intelligence services. Whether the Commerce Department holds durable authority to restrict software-as-a-service AI models globally remains legally untested and may not survive a direct challenge.
If this directive holds, the assumption that frontier models ship globally on launch day is no longer safe to make.
Frequently asked questions
What are Claude Fable 5 and Mythos 5?
Fable 5 is the publicly released, guardrailed version of Anthropic's Mythos architecture, aimed at applied and security research use cases. Mythos 5 is the full model, kept restricted to Project Glasswing partners. Both are now inaccessible following the Commerce Department directive.
Why did the US government block these models?
An unnamed company claimed it had demonstrated a jailbreak of Fable 5's safety guardrails, convincing the administration that unrestricted foreign access posed national security risks. The Trump administration had previously tried and failed to block the release before launch.
What does the export control order require?
The directive requires a license for any export, re-export, or transfer of Mythos 5 and Fable 5 outside the United States. It also restricts access for foreign nationals inside the country, making this one of the broadest controls ever applied to an AI model post-release.
What happens to developers already using Fable 5?
Anthropic has announced no migration path or restoration timeline. Customers on Azure and Amazon Bedrock lost access alongside direct API users.
About the Author
Guilherme A.
Former dentist (MD) from Brazil, 41 years old, husband, and AI enthusiast. In 2020, he transitioned from a decade-long career in dentistry to pursue his passion for technology, entrepreneurship, and helping others grow.
Connect on LinkedIn