U.S. Secretary of State Marco Rubio has instructed diplomats to challenge warnings that Washington can abruptly cut foreign users off from critical American technology, as recent restrictions on advanced artificial intelligence models fuel calls for greater digital independence in Europe and Asia.
A State Department cable dated July 16 and reviewed by Reuters supplied diplomatic posts around the world with arguments against the idea that the United States maintains a technology “kill switch.” The guidance followed a short-lived Trump administration measure that barred non-U.S. residents from accessing Anthropic’s most advanced AI models, Mythos and Fable, on national security grounds.
Anthropic responded to the restriction by suspending access to the models worldwide. The ban was withdrawn later in June, but the disruption gave foreign technology companies an opportunity to promote alternatives and strengthened the case made by some European policymakers for reducing dependence on U.S. platforms.
The cable did not identify Anthropic or explicitly discuss the temporary ban. It argued more broadly that narrowly targeted pauses and pre-release security reviews should not be interpreted as mechanisms for remotely disabling American products. Reuters reported that the State Department declined to comment on the cable and that the White House did not respond to a request for comment.
“There is no government ‘magic button,’” the cable said, describing the kill-switch characterization as exaggerated.
The dispute also reflects concerns about a June 2 executive order asking AI developers to submit models voluntarily for 30 days of cybersecurity testing before release. Such reviews are intended to identify dangerous capabilities, including the potential for powerful models to assist sophisticated cyberattacks. For governments and companies relying on American systems, however, the combination of testing requirements and temporary access restrictions has raised questions about whether U.S. policy could interrupt essential services with little notice.
Rubio’s guidance went beyond the immediate AI controversy. Diplomats were told to oppose what the cable called “digital sovereignty” initiatives when they restrict American companies’ market access, impose data or infrastructure localization requirements, levy network fees or require compliance with local content-moderation rules. Reuters previously reported that Rubio had issued similar instructions concerning data-sovereignty policies earlier in the year.
The cable also encouraged diplomats to counter arguments for “AI sovereignty,” including proposals for countries or regions to develop domestic alternatives to leading U.S. models. Its sales pitch emphasized the capabilities of American AI companies and presented building competing systems from scratch as an inefficient use of money and technical resources.
That position puts U.S. diplomacy at odds with a growing policy movement in Europe, where officials have sought more control over cloud computing, semiconductors, data and AI infrastructure. European concerns are not limited to technical reliability. They also encompass the possibility that access to dominant platforms could become leverage during a trade, security or diplomatic dispute.
European Parliament member Christophe Grudler said the Anthropic episode demonstrated that the United States could withdraw access to essential technology and was prepared to do so. Aura Salla, a member of the center-right European People’s Party, argued that Europe could not continue constructing its technology stack around services that a foreign government might abruptly make unavailable.
Edward Fishman, director of the Maurice R. Greenberg Center for Geoeconomics at the Council on Foreign Relations, told Reuters that the security rationale for restrictions on highly capable models was understandable because such systems can enhance complex hacking operations. He said the diplomatic message would nevertheless be difficult to sell after the administration used economic pressure, including tariffs and sanctions, against both adversaries and partners.
The cable’s central assurance was that customers can adopt U.S.-built AI infrastructure without surrendering ownership or operational control. Yet the temporary Anthropic ban showed that government policy can still affect who receives advanced models, even without a literal mechanism embedded in the technology. The task for U.S. diplomats is now to persuade allies that narrowly applied security controls will not become a broader source of economic or political leverage.
Sources: AI executive order