US orders Anthropic to cut foreign access to latest Claude models

Anthropic said the Trump administration has ordered it to block foreign nationals from using its newest artificial intelligence models, a move the company said will force a broader shutdown of access for customers. The order matters because it applies export-control rules to advanced AI software, limiting who can use tools built by a major US developer.

The San Francisco company, which makes the Claude chatbot, said Friday that the US government had cited national security powers in directing it to suspend access to Fable 5 and Mythos 5. Anthropic said the restriction covers foreign nationals inside and outside the United States, including foreign-national employees at the company.

Anthropic said the practical effect is that it must “abruptly disable” the two models for all customers to comply with the order. The company said access to its other models would continue.

Security concerns cited

Anthropic said the government did not provide detailed public reasoning beyond national security concerns. Semafor reported on June 14 that the order was linked in part to suspicions that a China-linked group had accessed Anthropic’s new model; Al Jazeera said it could not independently verify those reports.

David Sacks, an adviser to President Donald Trump, said on X that the government had received a warning that Fable 5 could be jailbroken, according to Semafor’s account. Sacks said Anthropic chief executive Dario Amodei was alerted and that the company did not fix the issue.

Anthropic disputed the basis for the order in its Friday blog post. The company said officials had given it only “verbal evidence of a potential narrow, non-universal jailbreak” and said it disagreed that such a finding justified pulling a commercial model used by hundreds of millions of people.

Kirsten Davies, the Pentagon’s chief information officer, backed the national security rationale in a June 13 post on X. Davies wrote that some concerns should take priority over revenue, clickbait and pre-IPO valuations, and ended the post with “America First. Always.”

Business and research impact

Anthropic released Fable 5 earlier in the week, based on its Mythos technology. Experts told Reuters that Mythos models could speed up advanced cyberattacks if misused, especially against sectors such as banking that depend on complex and aging technology systems.

Anthropic said it had worked with the US government and others on safety before launching Fable. The company also said rival AI models had shown similar abilities to find minor software bugs.

The order comes as the Trump administration has used export controls to restrict access to advanced technology, including chips sold by companies such as Nvidia and AMD, according to Al Jazeera. It also adds to tensions between Anthropic and the White House; Al Jazeera reported that Anthropic is suing the administration after being placed on a supply-chain blacklist over its refusal to let the US military use its models for domestic surveillance and fully autonomous weapons systems.

Al Jazeera reported that the limits could affect foreign research partners, H-1B visa workers in the United States and customers outside the country. The report also noted that companies such as S&P use Claude software to connect databases for financial advisers, investors and analysts.

Some technology figures criticized the measure. Kun Chen, described on X as a former senior engineer at Meta, Microsoft and Atlassian, said using foreign-national status to control access would be hard to enforce and easy for malicious actors to bypass. Zoho co-founder Sridhar Vembu said on X that the order showed technology’s central role in national security and sovereignty.

Anthropic said it believes there has been a misunderstanding and is working to restore access to the affected models as soon as possible.

This story draws on original reporting from Al Jazeera.