In a move that has no precedent in the history of commercial artificial intelligence, the United States government ordered Anthropic to immediately shut down access to its two most capable AI models — Claude Fable 5 and Claude Mythos 5 — citing national security concerns tied to a potential jailbreak. The directive came without warning, just three days after the models launched to widespread industry acclaim. The story that has unfolded since then is one of technology, national security, and a very public clash between a leading AI company and the government it has long worked closely with.
The Rise and Fall of Claude Fable 5
When Anthropic released Claude Fable 5 on June 9, 2026, it was hailed as the most capable AI model ever made widely available to the public. Sitting in a new class Anthropic calls “Mythos-class,” Fable 5 outperformed every competitor on nearly every benchmark the industry uses to measure intelligence — from software engineering and scientific research to visual reasoning and long-horizon autonomous tasks.
The numbers were extraordinary. Stripe reported that Fable 5 compressed months of engineering work into a single day, completing a codebase-wide migration of a 50-million-line Ruby codebase that would have taken an entire team more than two months. GitHub described it as opening up “a class of long-horizon problems that were out of reach for earlier models.” Legal tech firms found its contract redlines matched or beat senior lawyers in blind review. Hedge funds said it aced their trading analysis evaluations nearly across the board. One physics research firm said Fable 5 achieved in 36 hours what took GPT-5.5 four full days.
The model also came with an unprecedented safety architecture. Anthropic spent months red-teaming it with internal teams, the UK AI Safety Institute, and multiple private third-party organizations — thousands of hours of adversarial testing designed to prevent the model’s advanced cybersecurity capabilities from being weaponized. Anthropic was so confident in these safeguards that it publicly stated they were “the most robust of any model tested” by external partners, with zero successful universal jailbreaks found before launch.
Then, three days later, everything stopped.
The Government Steps In
At 5:21 PM Eastern Time on June 12, Anthropic received an export control directive from the US government. The order, citing national security authorities, instructed Anthropic to immediately suspend all access to Fable 5 and Mythos 5 for any foreign national — whether inside or outside the United States, including Anthropic’s own foreign national employees.
Because there was no practical way to implement such a restriction quickly without locking out all users, Anthropic made the difficult decision to disable both models entirely for every customer worldwide. Access to all other Claude models — Opus 4.8, Sonnet, and Haiku — remained unaffected.
In a public statement published the same day, Anthropic said the government’s letter did not provide specific details of its national security concern. The company’s understanding was that the government believed it had become aware of a method for bypassing, or “jailbreaking,” Fable 5. Anthropic reviewed what it believes was the basis of the directive and offered a stark rebuttal.
“The potential jailbreaks that have been disclosed to us are either entirely benign responses or are minor findings that provide no Mythos-specific uplift,” the company stated. “We have validated that the level of capability displayed there is widely available from other models, including OpenAI’s GPT-5.5, and is used every day by the defenders who keep systems safe.”
Anthropic chose to comply with the directive while publicly disagreeing with it — a calculated and unusually confrontational stance for a company that has prided itself on close collaboration with the US government on AI safety.
A Narrow Jailbreak, a Big Debate
At the heart of this controversy is a question that the AI safety community has debated for years: what counts as a dangerous jailbreak, and who gets to decide?
Anthropic has been clear about its position. The company acknowledged from the outset that perfect jailbreak resistance is not possible for any model provider. Instead, it adopted what it calls a “defense in depth” strategy — making jailbreaks either very narrow in scope or very expensive to produce, while using aggressive monitoring to detect and shut down successful attacks quickly. The company’s 30-day data retention policy for Mythos-class models, unpopular with many customers, was part of this defensive posture.
By Anthropic’s account, the jailbreak the government demonstrated was exactly the kind of narrow, non-universal case the company’s safety framework was designed to handle — one that exposed only “previously known, minor vulnerabilities” also accessible through other publicly available models. The company argued that applying the government’s apparent standard across the industry “would essentially halt all new model deployments for all frontier model providers.”
That argument puts Anthropic in an awkward position. The company has long been among the most vocal advocates for government oversight of powerful AI systems. It has actively supported the idea that regulators should have the authority to block unsafe AI deployments. But Anthropic drew a bright line around process: such oversight must be “transparent, fair, clear, and grounded in technical facts.” The June 12 directive, the company says, did not meet those standards.
What This Means for AI Development
The suspension of Fable 5 and Mythos 5 arrives at a particularly sensitive moment for Anthropic. The company filed confidentially for an IPO on June 1 — the first major AI lab to formally begin the public offering process — at a staggering valuation of $965 billion, following a $65 billion fundraising round. Its revenue run rate hit approximately $47 billion in May, up from roughly $10 billion the prior year.
The business interruption caused by the government’s directive has disrupted hundreds of millions of users globally and threatened relationships with major enterprise customers who had already integrated Fable 5 into their workflows. For a company weeks away from a historic IPO, the optics and operational damage are significant.
More broadly, the episode raises questions that extend well beyond Anthropic. If the government can direct a commercial AI company to disable its models globally — based on verbal evidence of a narrow potential vulnerability — what does that mean for the pace of AI development? What standards will govern future directives? And what happens to the millions of developers, researchers, and businesses who built systems around models that can vanish overnight?
The Fable 5 suspension is the first time a major commercial AI model has been pulled from global deployment by government order. It will not be the last conversation about where the line between national security and innovation should fall.
What Comes Next
Anthropic has stated it is working to restore access to Fable 5 and Mythos 5 as soon as possible and believes the situation represents a misunderstanding that can be resolved. The company has promised to share more technical details about the alleged jailbreak within 24 hours of the directive.
For users, developers, and AI watchers, the next few days will be telling. The outcome — whether Fable 5 is restored quickly or remains offline for a prolonged period — will shape how the entire industry thinks about government authority over AI deployment, and whether the current era of rapid model releases can continue at its current pace.
What’s clear is that artificial intelligence has arrived at a threshold its creators always said was coming: the point where the technology is powerful enough to attract not just users and investors, but the serious, binding attention of nation-states. The questions raised by Fable 5’s three-day run at the top of the AI world will define the next chapter of the industry.
Sources: Anthropic official announcements, CNBC, Al Jazeera, Fortune, NBC News
