Anthropic has been forced to take its newest artificial intelligence models offline worldwide following an emergency directive from the US government, escalating tensions between the AI company and federal regulators.
The move affects Fable 5 and Mythos 5, Anthropic’s latest and most advanced AI systems. While the government order was initially aimed at restricting access by foreign nationals, the company says technical and legal realities left it with no choice but to suspend the models for everyone.
The shutdown marks one of the most aggressive government interventions yet into the deployment of commercial AI technology.
Emergency Export Controls Trigger Global Shutdown
According to Anthropic, the US Commerce Department issued an emergency export control directive ordering the company to block foreign nationals from accessing Fable 5 and Mythos 5.
The definition of “foreign nationals” extends beyond people located outside the United States. It also includes non-US citizens living in the country and even some international employees working inside American companies.
Faced with those restrictions, Anthropic said it could not guarantee compliance while keeping the models operational.
As a result, both models were taken offline globally, including for US citizens and domestic customers.
Why Anthropic Could Not Enforce a Partial Ban
The company argues that implementing a selective restriction was not technically feasible.
One challenge involves real-time identity verification. Anthropic currently lacks a system capable of instantly determining the nationality or citizenship status of hundreds of millions of users accessing its services.
A second complication involves its own workforce.
Because the export restrictions also apply to foreign nationals employed by Anthropic, allowing the models to remain active internally could potentially expose the company to immediate regulatory violations.
To avoid that risk, the company opted for a complete shutdown.
Users attempting to access Anthropic’s platform can still use older models, but Fable 5 and Mythos 5 remain unavailable.
A Cybersecurity Dispute at the Center of the Controversy
The government’s concerns reportedly stem from the cybersecurity capabilities built into Anthropic’s latest AI architecture.
Federal officials fear advanced AI systems could potentially be used to identify software vulnerabilities that malicious actors might exploit against critical infrastructure, financial institutions, or government networks.
Those concerns intensified after regulators received reports of a successful “jailbreak”, a method used to bypass an AI model’s built-in safety restrictions.
Officials reportedly viewed the incident as evidence that advanced coding capabilities could be misused if the models remained widely accessible.
What the Reported Jailbreak Actually Did
According to technical details released over the weekend, the alleged jailbreak may have been far less dramatic than regulators initially feared.
Researchers reportedly used a sequence of carefully designed prompts to bypass certain safety controls within Mythos 5.
The result was not a destructive cyberattack or unauthorized system breach. Instead, the model analyzed a piece of software code, identified several security weaknesses, and suggested ways to fix them.
Supporters of Anthropic argue that this type of activity is common throughout the cybersecurity industry and forms the basis of many defensive security practices.
In other words, the model was performing tasks similar to those routinely carried out by security researchers and ethical hackers.
Anthropic Says Regulators Misunderstood the Risk
Anthropic has strongly criticized the shutdown, arguing that regulators overreacted to a limited technical finding.
The company maintains that the reported jailbreak was narrow in scope and did not demonstrate a widespread failure of the model’s safety systems.
Anthropic also argues that similar capabilities already exist across the AI industry.
According to the company, competing frontier models are capable of identifying software vulnerabilities without requiring complex workarounds or jailbreak techniques.
From Anthropic’s perspective, treating this incident as grounds for a full product recall could create a precedent that affects nearly every major AI developer.
Growing Friction Between Anthropic and Washington
The latest dispute comes amid broader tensions between Anthropic and US government agencies.
Relations reportedly became strained earlier this year after disagreements over potential military applications of AI technology.
Those tensions have now intensified as regulators and AI companies continue debating how powerful models should be deployed, monitored, and controlled.
The timing is especially significant because Anthropic is reportedly preparing for a major public offering, making any disruption to its flagship products particularly consequential.
What Happens Next?
For now, Fable 5 and Mythos 5 remain completely offline.
Anthropic’s legal and technical teams are reportedly working to negotiate with the Commerce Department while exploring ways to comply with the restrictions.
The company may eventually need to build new infrastructure capable of verifying user eligibility based on nationality and regulatory requirements.
Until then, customers will have to rely on Anthropic’s older AI models while the company and federal regulators attempt to resolve one of the most significant AI policy disputes seen to date.
The outcome could have implications far beyond a single company, potentially shaping how future AI systems are regulated, deployed, and secured across the industry.
