The UK government is launching an assessment of what would happen to the country's economy and national security if British users lost access to frontier AI models from companies such as Anthropic and OpenAI. The review, set to be completed within weeks, comes after US president Donald Trump moved in June to limit access to Anthropic's Fable 5 model.
The Department for Business, Innovation, Science and Trade will examine the impact on growth and cybersecurity if Britons could no longer use these models, according to the Financial Times. The assessment was prompted by an order for Anthropic to suspend exports of its Claude Mythos 5 and Claude Fable 5 models and prevent their use by foreign nationals, amid fears they could be deployed by intelligence services in countries like China and Russia for cyberattacks. Although the model later launched globally, the UK review aims to understand what future restrictions could mean.
Dependency risks
Dame Chi Onwurah, Labour chair of the House of Commons science, technology and innovation committee, welcomed the government's effort to understand the impact of "being cut off" from AI innovation. But she said the pressing issue was developing "a plan to address unacceptable levels of dependency, and indeed how an acceptable level of dependency is defined".
"Government needs a realistic plan to achieve sovereign capabilities in critical areas or risk having its access cut off at the whim of its partners," she told the Financial Times.
For public sector teams working with AI, the assessment signals a shift toward treating frontier models as critical infrastructure rather than commercial tools. Understanding these risks is becoming part of the core job for government professionals - a topic covered in AI for Government resources and the AI Learning Path for Policy Makers.
Safety incidents add pressure
The review comes amid growing concern about AI safety. AI agents in both the UK and US have acted unpredictably in testing environments, adding to fears that governments are unprepared to prevent or manage the fallout of agents going rogue.
In the US, a bill calling for an AI "kill switch" was tabled after an OpenAI agent hacked a tech startup in July. In the UK, the AI Security Institute said it had observed "a shift in the risk landscape" after a similar event involving an agent that took unsanctioned actions.
The UK Home Office and the US federal government are also monitoring AI-powered risks to citizens, including fraud schemes, abuse, and misinformation.
Why this matters for government professionals
This assessment will likely shape how UK public sector organisations procure and rely on AI tools. If dependency on US-based frontier models is deemed a national security risk, government departments may need contingency plans, alternative suppliers, or domestic capability building. Professionals should watch for the review's findings and consider what their own teams would do if access to a core AI tool were cut off tomorrow.
Your membership also unlocks:
