In an unprecedented move that sent shockwaves through the AI industry, Anthropic completely shut off access to its most advanced AI models — Mythos 5 and Fable 5 — on Friday night, just days after their much-anticipated launch. The abrupt shutdown came after the company received a US Commerce Department export control directive, citing national security authorities.
What the Government Directive Actually Says
The US government, citing national security authorities, issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States. This includes foreign national Anthropic employees, according to a statement posted online late Friday.
The directive effectively bars anyone who is not a US citizen or permanent resident from using or even accessing these models, regardless of their physical location. This is a significant escalation in how the US government treats advanced AI models as controlled technology.
Why Anthropic Had No Choice But to Pull the Plug
In a message posted Friday night, Anthropic said the only way for it to ensure compliance with that government order in the immediate term "is that we must abruptly disable Fable 5 and Mythos 5 for all our customers." The company did not have time to implement a more nuanced access control system, leaving a complete shutdown as the only viable option.
This decision affects not just international customers but also Anthropic's own workforce. Foreign national employees at Anthropic's US offices are now barred from accessing the very models they helped build.
The Jailbreak Concern That Triggered the Order
An Axios report cited an administration official saying that the administration is concerned by reports of a jailbreak that reportedly gets around broad classifier-based safeguards meant to block Fable 5 prompts. This suggests the government acted preemptively based on intelligence about potential security vulnerabilities in the model's safety systems.
The jailbreak reportedly bypasses the classifier-based safeguards that were designed to prevent the model from generating harmful or restricted content. If confirmed, this would represent a serious failure in Anthropic's safety architecture for its most advanced models.
Who Is Affected by This Shutdown
The immediate impact falls on Anthropic's customers who had just begun integrating Fable 5 and Mythos 5 into their workflows. These include enterprise clients, researchers, and developers who had been eagerly awaiting access to what Anthropic described as its most capable models yet.
For Anthropic's own foreign national employees — many of whom are top AI researchers — the situation is particularly jarring. They are now locked out of the very systems they developed, unable to continue their work on improving or maintaining these models.
Official Response from Anthropic and the Administration
Anthropic's statement emphasized that the company is complying fully with the directive while expressing frustration at the abrupt nature of the order. The company has not indicated whether it plans to challenge the directive legally or seek modifications.
The Trump administration official, speaking to Axios, framed the action as a necessary national security measure, pointing specifically to the jailbreak concern. The administration has not provided further details about the nature of the jailbreak or how it was discovered.
What This Means for AI Export Controls
This action represents one of the most aggressive uses of export control authority against a specific AI model deployment. Previously, export controls on AI have focused on chip exports and cloud computing access, not on directly ordering a company to disable access to its own software models.
The directive sets a precedent that the US government can, at any time, order an AI company to cut off access to its products based on national security concerns. This creates significant uncertainty for the entire AI industry about the stability of access to advanced models.
Confirmed Facts vs What Remains Unclear
Confirmed: Anthropic received a US Commerce Department export control directive Friday evening. The company disabled Fable 5 and Mythos 5 for all customers. The directive restricts access by foreign nationals anywhere in the world. Other Anthropic models remain unaffected.
Unclear: The exact nature of the reported jailbreak. Whether the jailbreak was actually successful or merely a theoretical concern. What specific national security risks the administration believes these models pose. Whether Anthropic will seek to negotiate a modified compliance approach.
Anthropic's Position in the AI Landscape
Anthropic has positioned itself as a safety-first AI company, often contrasting its approach with competitors like OpenAI. The company's focus on "constitutional AI" and classifier-based safeguards was central to the Fable 5 and Mythos 5 models. The reported jailbreak, if confirmed, would be a significant blow to this safety narrative.
The company's models are built on a foundation of careful safety testing and alignment research. This incident raises questions about whether even the most safety-conscious AI companies can fully anticipate or prevent misuse of their most powerful systems.
Risks and Balanced View
Supporters of the administration's action argue that the US cannot afford to let advanced AI capabilities spread to foreign adversaries, particularly if safety mechanisms can be bypassed. They see this as a necessary precaution in an era of heightened geopolitical competition over AI.
Critics, however, warn that this sets a dangerous precedent for government overreach into AI development. They argue that abrupt shutdowns harm legitimate research and business operations, and that the lack of transparency about the jailbreak concern makes it difficult to assess whether the response is proportionate.
Broader Pattern: US-China AI Competition Intensifies
This action fits into a broader pattern of escalating US government intervention in AI technology. The Biden administration previously imposed chip export controls, and the Trump administration has continued and expanded these efforts. The direct targeting of AI model access represents a new frontier in this competition.
Other AI companies are now watching closely. If the government can order Anthropic to shut down models, similar actions could be taken against OpenAI, Google DeepMind, or other frontier AI developers. This creates a new regulatory risk factor for the entire industry.
What Customers and Developers Should Do Now
For businesses and developers who were using or planning to use Fable 5 or Mythos 5, the immediate step is to assess alternatives. Anthropic's other models remain available, and competitors like OpenAI's GPT-4 or Google's Gemini may offer comparable capabilities.
Companies should also review their contracts with AI providers for force majeure or regulatory compliance clauses that might apply in such situations. Diversifying AI model dependencies may become a prudent strategy going forward.
What Happens Next
Anthropic is likely to seek clarification from the Commerce Department about whether a more targeted compliance approach is possible. The company may also explore legal options, though challenging a national security directive is difficult.
The administration may provide more details about the jailbreak concern, which could influence public and industry reaction. If the jailbreak is confirmed to be serious, it could justify the aggressive response. If it proves to be a false alarm, the backlash against government overreach could be significant.
Our Take
This incident reveals a fundamental tension at the heart of advanced AI development: the same capabilities that make these models powerful also make them potentially dangerous. The government's response, while drastic, reflects genuine concerns about AI safety in an era of geopolitical competition.
However, the lack of transparency about the specific threat and the abrupt nature of the shutdown raise legitimate questions about due process and proportionality. The AI industry needs clearer rules of the road — not just for export controls, but for how safety concerns are communicated and addressed before emergency measures become necessary.
For now, this episode serves as a stark reminder that the era of unfettered AI development is over. Governments are watching, and they are willing to pull the plug.
Frequently Asked Questions
Why did Anthropic shut down Fable 5 and Mythos 5?
Anthropic received a US Commerce Department export control directive Friday evening, citing national security authorities. The order required the company to suspend all access to these models by foreign nationals anywhere in the world. Anthropic said the only way to immediately comply was to disable both models for all customers.
What is the jailbreak concern mentioned by the administration?
An administration official told Axios that the government is concerned by reports of a jailbreak that reportedly gets around broad classifier-based safeguards meant to block Fable 5 prompts. The exact nature of this jailbreak has not been publicly detailed.
Are other Anthropic models affected?
No. Access to other Anthropic models, including earlier versions of Claude, is not affected by this directive. Only the newly launched Fable 5 and Mythos 5 models have been disabled.
Can foreign national Anthropic employees access these models?
No. The directive specifically bars access by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees. This means some of the researchers who helped build these models cannot access them.
