Anthropic did not make the decision to shut down Fable 5. The US government made it for them.
The United States government issued a national security legal order directing Anthropic to suspend access to its new Fable 5 and Mythos 5 AI models. The order came through an Export Control Directive, which is a legal mechanism that restricts or suspends the transfer of specific products, data, or technologies to foreign countries or to citizens of another country.
The catch here is significant. Because it would be essentially impossible for Anthropic to verify the foreign national status of every user in real time, the company had no practical choice but to cut off access for everyone. So what started as a directive targeting foreign nationals ended up taking down both models for all customers.
What the Order Actually Says
Anthropic shared the details publicly. The company announced that the US government, citing national security authorities, directed it to suspend all access to Fable 5 and Mythos 5 for any foreign national, whether inside or outside the United States. That includes foreign national Anthropic employees as well.
The order was received at 5:21 PM EST. Access to all other Anthropic models was not affected by the directive.
The government’s cited concern is that there is a way to bypass the safety guardrails built into Fable 5. That is the security issue at the center of all of this. We are talking about so-called jailbreaks, which are techniques that can be used to get an AI model to do things it is designed not to do.
Anthropic, for its part, does not agree that this rises to the level of a national security emergency.
Anthropic Pushes Back
Anthropic’s public response makes clear that the company believes it already has sufficient protections in place and that the government’s concern is overstated.
The company explained that it has instituted strong safeguards that greatly reduce the likelihood that Fable 5 is misused for tasks related to cybersecurity, among other areas. Anthropic noted that these safeguards are actually so extensive that many users have complained they are too broad, meaning the model was already being quite cautious about what it would and would not do.
On the specific question of jailbreaks, Anthropic acknowledged that perfect jailbreak resistance is not realistically achievable with today’s AI technology. But the company says that is precisely why it adopted what it calls a defense in depth strategy. The idea is to make jailbreaks either narrow in scope or very expensive to produce, and to back that up with thorough monitoring to detect and shut down any successful attacks quickly.
Anthropic’s position is that this approach reduces the risks posed by Fable 5 to a level comparable to risks that already exist with other models currently deployed across the AI industry. And critically, the company stated that it has not even received a disclosure of a concerning jailbreak that led to any harmful result. The potential jailbreaks it has been shown are described as either entirely benign responses or minor findings that do not provide any meaningful additional capability that was not already available.
The Broader Dispute
This did not happen in a vacuum. The US government and Anthropic have had an ongoing disagreement for some time, and this export control order appears to be the latest development in that conflict.
The dispute centers on Anthropic’s refusal to allow its products to be used for mass domestic surveillance and for fully autonomous weapon systems. We are talking about weapons that can independently identify targets and engage without a human making that final decision. Anthropic has consistently declined to support these use cases, and that refusal has put it at odds with certain government interests.
So the context here matters. This is not simply a story about an AI company having a safety gap. It is also a story about an AI company that has been unwilling to build tools for applications it considers ethically unacceptable, and is now facing legal pressure from the same government it has been pushing back against.
Public Reaction
The response on social media was fast and overwhelming. Anthropic posted on X that it believed the shutdown was a misunderstanding and was working to restore access as soon as possible. Within hours, that post had received over 7,000 replies and 32 million views. The reaction was largely sympathetic toward Anthropic and critical of the government’s action.
Some users pointed out that Fable 5 was already one of the most restricted AI models available. One user noted that the model’s safeguards were so extensive that basic biology questions and infrastructure-related tasks were being blocked. The argument being made was that Anthropic could not have done much more from a safety standpoint.
Others saw the shutdown as something bigger. Some users described it as marking the end of an era of open public access to frontier AI models, with concern about what happens when only governments and major AI labs have access to the most capable systems.
A significant portion of the response came from paying customers who had signed up specifically to use Fable 5. Several users described upgrading to more expensive subscription plans on the same day the model was shut down, and many asked for refunds. The timing made the situation particularly frustrating for people who had just committed to paying more for access they never got to use.
Not everyone placed the blame entirely on the government, though. Some users argued that Anthropic had spent years describing its most advanced models as extremely powerful and potentially dangerous, and that the government was essentially taking those warnings at face value. The argument was that you cannot market your AI as near-dangerous and then act surprised when regulators treat it accordingly.
What Happens Next
Anthropic has made it clear that it believes this is a misunderstanding and is actively working to get Fable 5 and Mythos 5 back online. If the company’s assessment of the jailbreak risk is correct, meaning the disclosed vulnerabilities are genuinely minor and do not provide any meaningful uplift to bad actors, then there is a reasonable path to a resolution where access is restored.
But that path runs through the government, and the existing tension between Anthropic and federal authorities adds complexity to how quickly and smoothly that resolution might come. Anthropic has been firm about not enabling mass surveillance or autonomous weapons, and it is not clear whether the export control directive and those prior disputes are entirely separate matters or connected in some way.
What is clear is that this situation has put a spotlight on the intersection of AI capabilities, national security policy, and commercial access in a way that the industry has not quite seen before. The question of who gets to decide when an AI model is too capable to be publicly available, and on what grounds, is now very much an open and active debate.
For users who lost access to Fable 5, the immediate concern is practical. Will service be restored, and if so, when? Anthropic has promised to work on it, but has not given a timeline. For the broader AI industry, the question is whether this sets any kind of precedent for how governments might intervene in the deployment of advanced models going forward.
Frequently Asked Questions
Q 1: What is the Export Control Directive that shut down Fable 5?
An Export Control Directive is a legal order issued by the US government that restricts or suspends the transfer of specific products, data, or technologies to foreign countries or citizens of another country. In this case, it was used to suspend foreign national access to Fable 5 and Mythos 5.
Q 2: Why did Anthropic shut down Fable 5 for all users if the order only targeted foreign nationals?
Because verifying the foreign national status of every user in real time is not practically possible, Anthropic had no choice but to disable both models for all customers to ensure it was in compliance with the directive.
Q 3: What is the government’s stated reason for the order?
The US government cited national security concerns, specifically that there is a way to bypass the safety guardrails built into Fable 5. The concern relates to potential jailbreaks that could allow the model to be misused.
Q 4: Does Anthropic agree with the government’s security assessment?
No. Anthropic disputes the government’s characterization. The company says it has not received evidence of a jailbreak that led to any harmful result, and describes the disclosed vulnerabilities as minor.
Q 5: What is a jailbreak in the context of AI models?
A jailbreak refers to a technique used to get an AI model to behave in ways it was designed to prevent, such as generating harmful content or bypassing safety restrictions. Anthropic acknowledged that perfect jailbreak resistance is not currently achievable, but says its defense in depth strategy makes misuse very difficult.
Q 6: What is Anthropic’s defense in depth strategy?
It is an approach that aims to make jailbreaks either narrow in scope or very costly to produce, while combining that with active monitoring to detect and shut down any successful attacks quickly.
Q 7: Are other Anthropic models affected by this order?
No. According to Anthropic, access to all other Anthropic models remains unaffected. Only Fable 5 and Mythos 5 were suspended.
Q 8: What is the broader dispute between Anthropic and the US government?
Anthropic has previously refused to allow its products to be used for mass domestic surveillance and for fully autonomous weapon systems, which are weapons capable of selecting and engaging targets without human involvement. That refusal has created ongoing tension with government authorities.
Q 9: Can customers get refunds?
Anthropic has not made a formal public statement on refunds as yet. Many users who recently upgraded their subscriptions specifically to access Fable 5 have publicly requested refunds.
Q 10: Is Anthropic working to restore access?
Yes. Anthropic posted on X that it believes the shutdown is a misunderstanding and is actively working to restore access as soon as possible, though no specific timeline has been provided.
