White House Asks OpenAI, Anthropic to Pause UK Model Access

The White House privately asked Anthropic and OpenAI to pause providing certain advanced AI models to Britain while U.S. officials assess potential risks and coordinate policy with U.K. counterparts. The request was made to executives at the two U.S.-based developers and was not announced publicly.

U.S. administration officials raised concerns that some powerful models could be used in ways that pose national security or public safety risks if shared without additional controls. The companies involved had been exploring commercial and government relationships in the U.K., and Washington sought a temporary pause so regulators and policymakers could evaluate technical safeguards, licensing options and whether existing export rules apply.

British officials were informed through diplomatic and industry channels and expressed surprise at the private nature of the request, according to people briefed on the exchanges. U.K. ministers and career diplomats have sought clarification about which models are affected and what criteria the U.S. will use to decide whether access can proceed. Several U.K. agencies that had planned trials or deployments of model-based tools for public services and research have paused engagements while they await further detail.

Anthropic and OpenAI have developed large language models and related systems that support a range of applications. Company representatives noted that models can be restricted or modified to reduce misuse, for example by limiting outputs, tightening access controls or adapting models for specific tasks. U.S. officials asked for time to determine whether such technical controls are sufficient, whether contractual or licensing terms are required, and how export-control frameworks should treat advanced AI capabilities with commercial and potential military uses.

Officials on both sides described the discussions as focused on risk assessment, user vetting and mechanisms to prevent re-export or misuse by third parties. No public ban or formal legal restriction has been announced; the request is described by officials as a temporary pause while policy and technical options are weighed.

The pause has caused delays for U.K. agencies planning to test new AI tools, creating uncertainty about project timelines and costs. Observers and analysts expect the talks to influence how U.S. companies work with allied governments and to prompt further work on licensing mechanisms or model-usage agreements.

Since large AI models became widely available, governments have debated how to manage harms while preserving benefits. The U.S. has explored export controls and new oversight for certain technologies and has engaged allies on harmonizing approaches. Industry has built safety layers and access controls while regulators seek clearer, consistent rules for cross-border use of advanced models.