Key Points:
- Anthropic abruptly disabled access globally to its newly launched Claude Fable 5 and Mythos 5 models.
- The mandatory shutdown complies with a U.S. government export control directive citing national security.
- The order bars any foreign national—including Anthropic’s own employees—from accessing the technology.
- Anthropic publicly disagreed with the decision, warning that it could halt all new deployments of frontier models.
U.S. Government Orders have forced artificial intelligence developer Anthropic to abruptly suspend all global access to its two most powerful, newly launched models, Claude Fable 5 and Claude Mythos 5. The sudden, mandatory shutdown responds directly to a highly restrictive export control directive citing “national security authorities.” Because the government’s order explicitly prohibits any foreign national—whether inside or outside the United States—from accessing the technology, Anthropic had to disable the systems for all customers worldwide to ensure full compliance. This unprecedented, sweeping intervention represents a major escalation in Washington’s effort to treat cutting-edge software code as highly classified military assets.
The federal directive arrived unexpectedly at 5:21 p.m. Eastern Time on Friday, catching both corporate executives and cloud partners entirely off guard. While the official letter did not provide detailed written specifics regarding the underlying national security threat, company officials believe the order stems from a narrow “jailbreak” or safeguard-bypass method that government researchers discovered. The verbal evidence provided to the company showed a user prompting Fable 5 to read a specific codebase and fix a small number of previously known, minor software flaws. While the administration viewed this capability as a potential cyber-weapons threat, the developer argued that other publicly available models can easily replicate the task without requiring a bypass.
The sudden regulatory recall comes just three days after Anthropic publicly launched the two sister models to widespread acclaim. Exceeding the capabilities of any previous AI system, both models were built on a massive default context window of 1 million tokens, with pricing set at a premium of $10 per million input tokens and $50 per million output tokens. While Fable 5 was designed for general commercial use with strict, built-in safety classifiers, Mythos 5 operated as an unsafeguarded variant reserved exclusively for a small group of vetted cyberdefenders and national security personnel under a highly confidential program called Project Glasswing.
The extreme regulatory anxiety surrounding the models is directly tied to their unmatched, highly autonomous capabilities in cybersecurity and coding. On industry-standard software engineering benchmarks, Fable 5 achieved a near-human score of 91 out of 100, far outpacing legacy models. During rigorous red-team testing before launch, the unsafeguarded Mythos 5 demonstrated the ability to discover and exploit zero-day vulnerabilities across every major operating system and web browser, fabricating working exploits in under a day. In the initial weeks of Project Glasswing, the model successfully identified over 10,000 high or critical vulnerabilities across systemically important software, prompting open-source maintainers to beg the developer to slow down disclosures because human engineers could not write security patches fast enough.
While fully complying with the government’s legal directive, the company’s leadership has publicly expressed strong disagreement with the recall’s heavy-handed nature. In an official statement, the firm warned that establishing a precedent in which minor, non-universal bypasses trigger the immediate shutdown of commercial models would effectively halt all frontier AI deployments across the industry. Chief Executive Officer Dario Amodei argued that while the government should have the statutory authority to block unsafe deployments, such decisions must follow a transparent, fair, and clear process grounded in objective technical facts rather than sudden, unilateral decrees.
The aggressive enforcement action marks the most significant step to date by the Trump administration to establish tight federal control over the domestic artificial intelligence sector. This direct intervention occurred just ten days after the president signed a new executive order establishing a voluntary federal framework to vet the national security risks of advanced models for up to a month before their public release. By leveraging export control laws to block foreign nationals—including Anthropic’s own foreign-born computer scientists—from accessing the software, Washington has signaled that it now views the weight of generative AI models as equivalent to that of nuclear secrets or advanced missile guidance systems.
This regulatory clash also occurs against a backdrop of long-standing friction between the technology startup and federal security officials. Earlier this year, the relationship between the company and the administration ruptured after executives refused to allow the U.S. military to use its Claude models to power fully autonomous weapons systems and run domestic surveillance programs. In retaliation, the federal government placed the developer on a supply chain blacklist slated to take effect later this year. The sudden shutdown of Fable 5 and Mythos 5 has reignited this feud, showing that the company remains a primary target of national security hawks in Washington.
The sudden, global disabling of Fable 5 and Mythos 5 serves as a stark warning to both the technology industry and international business partners. The comfortable assumption that advanced software algorithms could exist as borderless, universally accessible digital utilities has shattered against the reality of intense, superpower technological rivalry. While the company’s other models continue to operate normally, the complete withdrawal of its flagship software demonstrates that national security mandates will continue to override commercial interests. As the developer works to resolve the misunderstanding, the outcome of this dispute will dictate the future pace, safety, and ultimate sovereignty of the global artificial intelligence revolution.
