US asks Anthropic to block global access to top AI models: Why it matters

The US government has ordered AI company Anthropic to block all foreign nationals from accessing its latest AI models, Fable 5 and Mythos 5, citing national security concerns. This directive, issued under Donald Trump's administration, aims to prevent advanced AI technology from falling into the wrong hands, with reports suggesting suspicions of a China-linked group accessing the model. Anthropic stated the government provided "verbal evidence of a potential narrow, non-universal jailbreak," which the company believes does not warrant recalling a widely deployed commercial model. This action is part of a broader US policy of export controls on high-tech goods. The order will require Anthropic to disable these specific models for all customers to ensure compliance, though access to other Anthropic models remains unaffected.

AI company Anthropic will disable its advanced Fable 5 and Mythos 5 models for all users following a US government order to suspend foreign national access. The directive, citing national security concerns, stems from the government's belief that safeguards on Fable 5 can be bypassed to identify software vulnerabilities. Anthropic stated it received verbal evidence of a potential narrow jailbreak and disagrees that this warrants disabling the commercial models. This action represents an escalation in US efforts to restrict foreign access to AI capabilities, moving beyond controls on hardware. Anthropic aims to restore access as soon as possible, believing there is a misunderstanding.

The US government has ordered AI firm Anthropic to suspend access to its Fable 5 and Mythos 5 models for all foreign nationals, citing national security concerns. Anthropic received the export control directive on Friday and immediately blocked access for affected users, including foreigners currently in the US and those working at the company. The Mythos 5 model is noted for its ability to detect software vulnerabilities, a capability that could be a concern if it falls into the wrong hands. The Fable 5 model, based on Mythos technology, has its cybersecurity and biotechnology capabilities blocked. Anthropic stated it received limited details regarding the specific security concerns from the government.

The U.S. government has directed Anthropic, a prominent AI company, to prevent foreign nationals from accessing its newest AI models, Fable 5 and Mythos 5. This directive, issued by Washington, cites national security concerns as the reason for the restriction. The measure is an unprecedented export control specifically targeting AI models and was implemented shortly after Anthropic released them. According to Anthropic, the order arrived on Friday evening and applies even to the company's foreign-born employees. This action signifies Washington's growing involvement in regulating advanced artificial intelligence technologies.

AI company Anthropic has released Fable 5, a publicly accessible version of its advanced Mythos AI model, while restricting its use in sensitive areas. This model, previously limited to partner institutions due to cybersecurity concerns, is now available for tasks like software coding and research. An unrestricted version, Claude Mythos 5, remains accessible to select organizations, including those in Anthropic's Project Glasswing cybersecurity program. Anthropic has implemented safeguards, routing queries about critical infrastructure or sensitive scientific topics to a less powerful model, and has undergone extensive "red-teaming" and bug bounty programs to test its security. The release comes with a higher price point than previous models, and Anthropic is also leasing data center space from Elon Musk's xAI.

Anthropic has released a public version of its powerful Mythos AI model, now called Claude Fable 5. This is Anthropic's most capable model intended for broader use, demonstrating strong performance in software engineering and analytics. Access to Mythos was previously restricted to approximately 200 organizations, including the US government through the Glasswing program, after an earlier preview revealed its significant ability to identify software flaws. To mitigate risks, Anthropic has implemented safeguards that prohibit the model's use in sensitive areas like cybersecurity. The company announced this rollout on Tuesday.