Anthropic has suspended access to its most advanced AI models, Fable 5 and Mythos 5, following a directive from the US government citing national security concerns. The government ordered the company to block these models for all foreign nationals due to fears that a potential jailbreak (a method to bypass built-in safety restrictions) could allow the software to identify vulnerabilities in code. Anthropic stated it received only verbal evidence regarding this narrow, non-universal risk and disagreed that such a finding warrants a recall of commercial models currently deployed to hundreds of millions of users.

This move represents a significant escalation in US efforts to limit foreign access to AI capabilities, shifting focus from restricting the hardware used to power AI toward restricting access to the models themselves. Anthropic reported that the government previously pressured the company regarding the use of its models for domestic surveillance and fully autonomous weapons systems, leading to a proposed supply chain blacklist earlier this year. The current suspension forces a total service disablement of Fable 5 and Mythos 5 for all customers to ensure regulatory compliance.

Kirsten Davies, the Pentagon's chief information officer, confirmed the Department of Defence's support for the order, stating that national security takes precedence over revenue and valuation metrics. Anthropic, which filed for a confidential IPO last month, defended its safety protocols, noting that its models have guardrails against risky cybersecurity applications that some users already criticize as overly broad. The company warned that applying this standard across the industry would effectively halt all new model deployments for frontier AI developers, and it is currently working to resolve the issue to restore service as soon as possible.

Anthropic has suspended access to its most advanced AI models, Fable 5 and Mythos 5, following a directive from the US government citing national security concerns. The government ordered the company to block these models for all foreign nationals due to fears that a potential jailbreak (a method to bypass built-in safety restrictions) could allow the software to identify vulnerabilities in code. Anthropic stated it received only verbal evidence regarding this narrow, non-universal risk and disagreed that such a finding warrants a recall of commercial models currently deployed to hundreds of millions of users.

This move represents a significant escalation in US efforts to limit foreign access to AI capabilities, shifting focus from restricting the hardware used to power AI toward restricting access to the models themselves. Anthropic reported that the government previously pressured the company regarding the use of its models for domestic surveillance and fully autonomous weapons systems, leading to a proposed supply chain blacklist earlier this year. The current suspension forces a total service disablement of Fable 5 and Mythos 5 for all customers to ensure regulatory compliance.

Kirsten Davies, the Pentagon's chief information officer, confirmed the Department of Defence's support for the order, stating that national security takes precedence over revenue and valuation metrics. Anthropic, which filed for a confidential IPO last month, defended its safety protocols, noting that its models have guardrails against risky cybersecurity applications that some users already criticize as overly broad. The company warned that applying this standard across the industry would effectively halt all new model deployments for frontier AI developers, and it is currently working to resolve the issue to restore service as soon as possible.