Anthropic announced on Friday that it will immediately cut off access to its most advanced AI systems for foreign users after receiving a directive from the US government tied to national security concerns. According to the company, officials ordered it to suspend access to its latest models — Fable 5 and Mythos 5 — for all non-US nationals. Anthropic said the government did not provide detailed evidence explaining the security risks behind the decision. The AI firm said authorities appear concerned that users may have discovered a way to bypass, or “jailbreak,” safeguards designed to stop Fable 5 from being used to detect software vulnerabilities.

The move comes at a sensitive moment for Anthropic, which had recently shown signs of improving its strained relationship with the US government. Tensions had escalated earlier this year after the company reportedly refused to allow its AI systems to be used for domestic surveillance or fully autonomous military weapons. Following that dispute, Anthropic was placed on a federal supply-chain blacklist scheduled to take effect later this year. The latest directive also signals a significant expansion of US efforts to control advanced artificial intelligence technologies. Until now, Washington’s restrictions have largely focused on limiting exports of AI chips and hardware rather than blocking direct access to AI models themselves.

Anthropic argued that the government’s concerns were based only on “verbal evidence” of what it described as a limited and non-universal jailbreak issue. “We do not believe that a narrow theoretical vulnerability justifies pulling a commercial AI system used by hundreds of millions of people,” the company said. The dispute highlights growing disagreements between regulators and AI developers over how serious model “jailbreaks” actually are and how governments should respond to them.

Ironically, only days earlier, Anthropic had publicly supported stronger AI oversight in the United States, including measures allowing authorities to block dangerous AI systems. However, the company criticized Friday’s action as lacking transparency and proper evidence-based regulation. Pentagon Chief Information Officer Kirsten Davies defended the administration’s approach in a post on X, saying national security must take priority over commercial interests. “Some things matter more than revenue cycles, clickbait headlines, and IPO ambitions. America First. Always,” Davies wrote.

Anthropic recently filed confidential paperwork for a US initial public offering, placing it ahead of competitor OpenAI in the race toward entering public markets. The company introduced Claude Fable 5, part of a new generation of AI systems it refers to as “Mythos-class” models. These systems include built-in restrictions aimed at preventing misuse in sensitive areas such as cybersecurity, though some users have criticized those limitations as being too restrictive. Security analysts have warned that powerful AI models in the wrong hands could increase the sophistication of cyberattacks, particularly against industries like banking and critical infrastructure that depend on aging and interconnected digital systems.

Anthropic said it had coordinated with US authorities before launching the new models and argued that competing AI systems from other companies demonstrated similar capabilities in identifying small coding vulnerabilities. The company confirmed that, to comply with the government order, it must now disable access to Fable 5 and Mythos 5 worldwide for foreign nationals. Other Anthropic AI products will remain available. Anthropic added that it believes the situation stems from a misunderstanding and said it is working with officials in hopes of restoring access quickly.

“If this standard were applied industry-wide, it could effectively stop future launches of advanced AI models across the sector,” the company warned. Amazon Web Services later stated that Anthropic had requested the removal of access to the models for all users globally. A US official confirmed that the Commerce Department issued the export control directive restricting access to Fable 5 and Mythos 5 for foreign nationals. Dean Ball, a former White House adviser involved in drafting the administration’s AI Action Plan in 2025, suggested on X that the policy could require users to prove American citizenship before using Anthropic’s newest models — even if they are physically located in the United States. The situation has also raised questions internally because several prominent Anthropic employees, including co-founder Chris Olah, researcher Andrej Karpathy, and philosopher Amanda Askell, were born outside the US. The company declined to comment on whether foreign-born staff members would also face restrictions.

Disclaimer: This image is taken from Business Standard.