The UK government is assessing the potential economic and security impact of losing access to advanced artificial intelligence (AI) models developed by companies such as Anthropic and OpenAI, according to the Financial Times.
The Cabinet Office has asked the Department for Business, Innovation, Science and Trade to examine how restrictions on access to frontier AI could affect productivity, research, cybersecurity, software development, and business competitiveness.
The review follows the US government’s temporary restrictions on foreign access to Anthropic’s advanced AI systems earlier in June 2026. It reflects concerns over the UK’s reliance on overseas companies for access to increasingly advanced AI technology.
In June 2026, the Trump administration issued a directive on export control and ordered Anthropic to stop access to its advanced AI models as a security measure.
The US Department of Commerce issued the order on 12 June 2026, requiring access to be blocked for foreign nationals, both inside and outside the country, including Anthropic staff who are not citizens of the United States.
Because Anthropic said it could not reliably distinguish foreign from domestic users in real time, the company later that day disabled Fable 5 and Mythos 5 for all customers globally, effectively shutting down the models for nearly three weeks.
At the end of June 2026, the Department of Commerce informed Anthropic that the export restrictions on Fable 5 and Mythos 5 were removed, which enabled the company to unblock access to its AI systems. Anthropic confirmed the removal of these restrictions in its press release and social media on 30 June, leading to the end of the 19-day suspension.
The Home Office and UK economic security officials are conducting a coordinated assessment of advanced AI as a dual-sided risk: the national security threat if hostile or criminal actors use cutting-edge models to identify and exploit unpatched cyber vulnerabilities, and the economic cost if British companies lag behind in adopting frontier AI.
The UK Government Resilience Action Plan 2026 Implementation report dated 14 July 2026 mentions that AI presents “not only a number of advantageous opportunities but also vast and overlapping threats” to the economies, jobs, and trust of the society of the UK. The report also states that NSRA is updated on an ongoing basis to keep up with the changes in the threats related to AI.
At the same time, the National Cyber Security Centre’s report “Impact of AI on cyber threat from now to 2027” warns that by 2027 AI tools will “almost certainly enhance threat actors’ capability to exploit known vulnerabilities,” increasing attacks on systems that have not been patched, while more skilled actors could use AI to improve zero-day discovery and exploitation.
In addition, the National Risk Register 2026, which lists 95 acute risks overall (87 standalone risks and 8 connected scenarios), was published by the UK Cabinet Office. Critical infrastructure is exposed to new AI-driven cyber dangers when static national planning is replaced with a dynamic resilience architecture. It creates a category called “digital resilience failure”, clearly connecting developments in AI to increased risks to vital services and data networks.
Dame Chi Onwurah, the Labour chair of the House of Commons Science, Technology and Innovation Committee, said to the Financial Times the government’s review of the potential impact of losing access to advanced AI systems was a positive step. However, she said ministers also need to define what constitutes an acceptable level of technological dependence and develop plans to address risks where that dependence becomes excessive.
Onwurah said her committee has repeatedly urged the government to establish a realistic strategy for developing sovereign capabilities in critical technologies. Without such a plan, she warned, the UK risks having access to key technologies restricted by decisions made by its international partners.
Ciaran Martin, former chief executive of the National Cyber Security Centre, explained to the Financial Times that the experience with export controls on Anthropic’s Fable 5 model was reassuring in some respects.
Martin said to FT, “What was the consequence? Anthropic took the whole thing off the market, as it couldn’t meaningfully distinguish between nationals and foreigners.”
The UK government has committed £500 million through its SovAI Fund and announced plans to invest £1.1 billion in AI computing infrastructure in April 2026. However, building alternatives to leading frontier AI labs requires time and resources. Developing advanced models, computing capacity, semiconductor supply chains, data centres, and skilled talent networks is a long-term process.
Britain could consider regulating advanced AI models if its current voluntary testing system no longer provides adequate public protection, AI Minister Kanishka Narayan told Reuters.
Following the AI Safety Summit of 2023, the AI Security Institute in the UK has been granted voluntary pre-deployment access to the AI models of firms including OpenAI, Anthropic, and Google. This enables the institute to conduct an assessment of capabilities and potential risks of AI systems before they are rolled out.
Narayan told Reuters that the institute has received pre-deployment access to nearly every frontier AI model developed by Western companies, making the UK the only country outside the United States with that level of access.
He described the position as unique and said the government’s priority remains the protection of the public. According to Narayan, policymakers are focused on outcomes rather than on any regulatory approach.
