U.S. Blocks Anthropic AI Access

Anthropic said it has disabled access to its Fable 5 and Mythos 5 AI models after receiving a U.S. government export-control directive that it says was based on national security authorities. The company said it was ordered at 5:21 p.m. ET on Friday to suspend access to the models for “any foreign national,” including foreign-national employees, and it responded by cutting off the models for all customers to ensure compliance. Anthropic said its other models are unaffected. The move came only days after Anthropic publicly launched the two models and promoted them as highly capable, with Fable 5 described as the company’s first advanced public release following new safeguards aimed at restricting high-risk responses. The article notes that the models were built on Claude Mythos Preview, which had already drawn attention for strong cybersecurity capabilities and was being limited to a small group of companies through Project Glasswing rather than broadly released. Anthropic said the government did not explain the specific national security concern, and the company apologized to customers for the disruption while arguing that such restrictions should follow a transparent, fair, and technically grounded process. The article also places the decision in the context of Anthropic’s broader tensions with the U.S. government, including an earlier dispute with the Department of Defense in which the DOD labeled Anthropic a supply chain risk. Anthropic sued the Trump administration over that blacklisting, and the case is still ongoing.

Anthropic said the U.S. government ordered it to suspend access to its new Fable 5 and Mythos 5 AI models for all foreign nationals, including foreign employees of the company, in the name of national security. The move, reportedly issued by the Commerce Department, is unusual because it does not merely restrict exports of technology to certain countries; it dictates who may use a company’s AI systems. The article frames the decision as a major escalation in the federal government’s oversight of advanced AI and another sign of a tense relationship between Anthropic and the Trump administration. Anthropic had recently limited public access to Mythos and released Fable with guardrails meant to reduce misuse in cybersecurity, biology, and other sensitive areas. The company said the government’s order was a misunderstanding and that it was trying to restore access. The restrictions could slow Anthropic’s development and complicate work by researchers and employees from allied countries. The article also places the episode in a broader policy context, noting that the White House has begun considering stronger oversight of AI systems and that rival OpenAI has also imposed limited release controls on its own cybersecurity-focused model.

The article reports that Washington has ordered Anthropic to prevent foreign nationals, including some of its own foreign-born employees, from accessing its newest AI models, Fable 5 and Mythos 5, due to national security concerns. According to Anthropic, the directive was issued abruptly on Friday evening and required the company to shut down access to those top-tier systems for all customers in order to comply. The move is described as an unprecedented export-control action because it is the first known U.S. restriction aimed at specific AI models rather than broader technologies or companies. The timing is notable because the order came only days after Anthropic launched the models, which the company had promoted as its most powerful to date. Anthropic had introduced Fable 5 and Mythos 5 on Tuesday. Fable 5 was publicly released with added safety and security restrictions, while Mythos 5 was limited to trusted partners. The company had already warned that Fable 5 could be abused to cause serious harm if those safeguards were removed, especially in sensitive areas such as cybersecurity, biology, and chemistry. The article also notes that Mythos 5 had previously been previewed to a small set of industry partners under Project Glasswing, with claims that it offered exceptional cybersecurity capabilities. Overall, the article highlights the growing willingness of the U.S. government to intervene directly in frontier AI development when it believes access could create security risks, and it shows how quickly new AI products can become subject to export-style restrictions.

Anthropic said it would abruptly disable its most advanced AI models, Fable 5 and Mythos 5, for all users after receiving a US government export-control directive that effectively bars foreign nationals from accessing them. The company said the government cited national security concerns and a possible “jailbreak” method that could let users bypass safeguards, but Anthropic said it was given only limited verbal evidence and no detailed, fact-based explanation. The move marks a sharp escalation in US efforts to restrict access to frontier AI systems, extending export controls beyond chips and hardware to the models themselves. The dispute reflects mounting tension between AI developers and regulators over how to handle model risks, especially in cybersecurity. Anthropic said its new Mythos-class models include guardrails against high-risk uses, such as identifying software vulnerabilities, though some users have criticized those safeguards as overly broad. Experts warned that in the wrong hands, these models could enable sophisticated cyberattacks, particularly against sectors such as banking. Anthropic argued the government’s action was disproportionate, saying a narrow, non-universal jailbreak should not justify pulling a commercial model used by hundreds of millions of people. The order also comes amid a fraught relationship between Anthropic and the Trump administration, including earlier clashes over military use of AI and a supply-chain blacklist. Anthropic said it is working to restore access and believes the government may have misunderstood the risk, while a US official confirmed the directive applies to foreign nationals. The episode highlights growing uncertainty over who may use advanced US AI models and whether citizenship checks could become necessary for access.