As the celebrations of American independence over the summer should have reminded us, Britain’s relationship with the United States has often been more difficult than we might like to think. Nonetheless, over the past 18 months, the bond has come under unusual strain.
Last summer’s Strategic Defence Review emphasized the upsides of the alliance but acknowledged America’s change in security priorities, ‘as its focus turns to the Indo-Pacific and the protection of its homeland’, and that, consequently, ‘the UK’s longstanding assumptions about global power balances and structures are no longer certain’.
That was before President Donald Trump threatened to invade a NATO ally and sought to press Britain to join his misbegotten war with Iran. As General Sir Richard Barrons, one of the review’s authors, puts it: ‘What used to feel like a beneficial alliance with a friendly big brother now feels like a transactional alliance with a genuine Big Brother.’
Worse, this fissure has appeared just as Britain is facing significant threats from an increasingly hostile Russia, is still working to repair its relationships with Europe after Brexit and is heavily burdened with debt. This confluence of crises is increasing pressure on the British state to shore up national security and defence, while making that aim more difficult to achieve.
At the sharpest point of this dilemma sits artificial intelligence. As the technology rapidly develops, it is becoming more central to our defence and security capabilities – particularly in operational planning and decision-making, predictive analysis and the use of uncrewed systems – but its development is being led by our ever-less-reliable ally across the Atlantic.
Outsourcing defence
AI may be helping to make weaponry more autonomous, but what about the autonomy of the states deploying it? Since DeepMind was acquired by Google in 2014, Britain has developed nothing approaching the vast, cash-ravenous general-purpose foundation models pioneered by American companies such as OpenAI and Anthropic.
These models are powerful and versatile, enabling the creation of generative AI applications, capable of complex, multistep tasks. Following the first cyber-attacks launched in July by an autonomous AI, these applications’ rapid development has triggered increased concern about a catastrophic loss of human control – a problem the US and its Chinese rivals need to confront.
But British reliance on these systems presents a more immediate risk: exposure to a ‘kill switch’. The Trump administration’s recent intervention ordering Anthropic to block foreign nationals from using its Claude Fable 5 programme was motivated by these rising safety concerns, but it also exposed the profound impact Washington can impose on any customer of a US company.
Beyond AI, a similar ‘shutter control’ risk might undermine British defence capabilities that rely on innovative US technology. As one former senior British military officer notes, Washington could choose to refuse software upgrades for the Royal Air Force’s 48 F35B fighter jets, for example. A US company such as Elon Musk’s satellite firm Starlink might cut off access to its services, on which the Ministry of Defence has reportedly spent more than £16 million in its efforts to back Ukraine’s defence against Russian aggression.
Britain reliant on US tech
Yet currently, as the investor Angus Hanton, author of Vassal State: How America Runs Britain, points out, Britain is deepening its dependence on America, not least by allowing its defence companies to go the way of DeepMind and be bought up by US firms.
Between 2020 and 2022, Britain permitted two of its leading defence firms – the aerospace manufacturer Cobham and the components company Meggitt – to be bought by US companies. More recently, the precision instruments supplier Spectris and the cyber security firm Darktrace have gone the same way.
Meanwhile, the British state has become increasingly reliant on US tech companies, such as Microsoft and Amazon Web Services – and, most controversially, Palantir, which was recently awarded a three-year, £240 million contract without competitive tendering by the Ministry of Defence for ‘support to data analytics capabilities supporting critical strategic, tactical and live operational decision making’. In June, a parliamentary select committee concluded that for Britain, ‘Palantir’s increasing presence across the public sector represents an unacceptable point of weakness.’
On the European continent, once-unthinkable moves are afoot to reduce technological dependence on America. Many Danish government systems, for example, now run on a system built by a domestic company, rather than US ones. The German military will reportedly not be contracting Palantir for the moment.
In this context, those charged with maintaining British national security must decide how far they need to reduce its dependence on the US, and which elements of the relationship must be preserved. As national security-focused AI develops at speed, particularly in the US, can Britain steer a path between ceding ever more control to its powerful big brother, and trying to bolster national independence by relying on more limited tech?
Home-grown technology
One potential approach that emerges from conversations with former senior military figures and AI analysts is to acquire the tech necessary to keep Britain safe from a range of sources, chosen according to the degree of complexity and security risk involved.
For simple defence operations that do not carry security implications, foreign technology solutions could be bought off-the-shelf at little risk. The steep energy costs of running AI could be met in part by buying cheap Chinese solar panels, for example. This poses no real problem as they are long-lasting, software-free purchases that require little maintenance.
Where national security is a factor, Britain could avoid simply defaulting to reliance on the big US foundation AI models and turn instead to smaller specialist applications developed domestically. Barrons, a former commander of the UK’s Joint Forces Command and a senior consulting fellow at Chatham House, argues that for command-and-control systems, for example, maintaining sovereign control outweighs the ‘potential degradation in performance’ from losing access to the most advanced US systems, particularly where the tasks concerned do not require generative AI.
Britain has a wealth of leading AI technologists, and efforts are underway to make it easier for them to develop domestic applications. With the backing of the government’s Sovereign AI Fund, the UK-based AI lab Cosine is currently developing what it says will be ‘Britain’s first fully sovereign frontier AI model’. Lumen Sovereign is intended to be ‘an AI capability that operates with zero dependence on foreign infrastructure’. This will reportedly be built using models trained on less data than US equivalents, which will be sufficient for specialized requirements.
As some British AI analysts point out, smaller specialized applications are not necessarily a second-best option. The big generalist models are foreign-controlled and increasingly expensive as token costs increase. Smaller, ‘open weight’ models can give greater control and be run at lower cost. One overall problem, however, is that generative AI can be unreliable. A given set of inputs won’t always produce the same outputs – which is problematic when flying a plane autonomously, for example.
This approach – diversifying supply and reducing dependence on any one provider – is similar to the strategy being advanced by the European Commission. Under its proposed Cloud and AI Development Act, tasks that require AI would be categorized by sensitivity, with the most sensitive being allocated to a ‘full sovereign European stack’.
Beyond AI, this approach could be applied to other frontier technologies, such as robotics, quantum computing and graphene, by fostering the creation of home-grown startups that draw on the UK’s strengths in scientific innovation. Britain has demonstrated the worth of its military tech capabilities in its contribution to Ukraine’s astonishingly effective shift into large-scale drone production. As the former Chief of the Defence Staff General Sir Nicholas Carter has argued, innovations in weaponry require the evolution of a new way of warfare to match them both to the threats Britain faces, and to budgetary constraints.
However, there are various measures the UK would also need to take to bolster its autonomy in this field. Barrons, for example, suggests it would require the construction of sovereign data centres, though this would need to win public consent, given the growing opposition to such centres in America.
Upsetting the elephant
The economist Will Hutton, meanwhile, suggests that nervousness about American hostility to the UK blocking takeovers of its innovative start-ups too easily trumps the need to bolster the UK’s autonomy. ‘At some stage’, he suggests, government must ‘judge when you can upset the elephant.’
Britain also needs to develop greater self-sufficiency at more basic levels that underpin new technology: the energy that AI consumes so voraciously, and the processing of critical minerals needed to build components. ‘Rare earths’ themselves are not rare, but the capacity to make them usable is. China currently dominates that process.
American efforts to bolster its position in this area have focused on the large-scale processing of metals such as copper, zinc and tin. Here too, there may be an opportunity for Britain to develop bespoke capacity, reviving small-scale processing of high-value metals such as tungsten.
Finally, government would need to overcome Whitehall’s one-sided assessment of risk, which focuses on the measurable costs of action rather than on the harder-to-measure costs of inaction, particularly in procurement and speculative investment in potentially useful start-ups.
In 2025, prompted by the Strategic Defence Review, the MoD launched UK Defence Innovation, a team ‘working to break down traditional barriers’ to accelerate the creation of cutting-edge capabilities. The new prime minister, Andy Burnham, has promised re-industrialization, higher defence spending and a reining in of Treasury dominance. Whether this will be sufficient to shore up UK sovereignty remains to be seen.
Efforts to reduce dependence could clearly be enhanced by co-operation with other middle powers facing similar challenges. Britain has strengthened links with European states through the Joint Expeditionary Force to Ukraine, and its Security and Defence Partnership with the European Union; it is planning to develop ‘deep strike’ missiles in co-operation with Germany under the Trinity House agreement.
However, it has not managed to access the EU SAFE programme which helps finance collaborative defence procurement – something Olivia O’Sullivan, director of Chatham House’s UK in the World Programme, has identified as a priority for Burnham. As a former senior military officer notes, addressing the gaps left by US withdrawal will require Britain and its European allies ‘to suspend national parochialism when it comes to their defence industries’.
Beyond Europe, replacing US capability would require stronger relationships and the pooling of talent and capital with other middle powers, alongside greater efforts to achieve interoperability within NATO. Rana Mitter of the Harvard Kennedy School draws attention to the under-recognized potential of working with Japan, given its deep capacity in AI, including in capital and research and development, its wariness of China and nervousness about US unpredictability, and its warmer relations with the UK. In this vein, Britain has recently announced a UK–Japan Frontier Technology Partnership.
Maximizing British and middle power independence evidently has its limits. Where foundation models are necessary, Europe has not invested in research and development early enough to be able to compete with US or Chinese models, France’s Mistral model notwithstanding. For the moment, this leaves using the US models as the only option. As the Strategic Defence Review underlines, transatlantic links between officials remain cordial, as they do between the British government and US defence businesses. And a senior military officer suggests that it would still be preferable to keep the US in NATO for the credibility of Article 5’s mutual defence clause, though this might require some ‘constructive ambiguity’.
Strength through small gains
In the context of this more transactional but still functional relationship, British ministers should not underestimate the cards they have available to play. These include GPS ground stations in the Indian Ocean and the Atlantic, signal intelligence localities in Cyprus and the UK’s widely admired special forces. Then there is Britain’s growing strength in computing power, and the depth and closeness of its defence relationship with Ukraine, including a new data-exchange deal which allows British access to Ukraine’s Avengers AI Labs. This could help Britain to sharpen its AI application performance in ways that improve its hand in its relationship with the US.