What Liz Kendall’s AI sovereignty speech means for Britain

If ‘AI sovereignty’ sounds like the sort of phrase made for Whitehall seminars, Liz Kendall’s speech at RUSI tries to bring it down to earth. In the version published on GOV.UK, she argues that AI is no longer a side story about chatbots or office tools. It is becoming part of how countries make money, defend themselves and decide who gets a say in the future. Her starting point is that two big changes are happening at once. The old global order is weaker than it was, and new technology is moving with startling speed. In her telling, those changes push each other forward: rivalry between states drives investment in chips, drones and data systems, while online platforms and algorithms help deepen political division.

Kendall then makes the speech’s biggest claim: AI is the defining currency of this period. She says earlier generations were shaped by navies, railways and power grids; now the big prize is control over powerful AI systems. That matters, she argues, because the same tools that may help discover medicines or new materials may also strengthen militaries and cyber operations. She points to the speed of change. According to the speech, frontier models have moved from toddler-level tasks to work beyond PhD level in about seven years, and capability is now doubling every four months rather than every eight. She also warns that around 70 per cent of global AI compute is controlled by just five companies. **What this means:** when ministers say ‘compute’, they mean the specialist chips, data centres and electricity needed to train and run advanced AI. If only a few firms control most of that, governments risk depending on decisions they do not make themselves.

That is where ‘AI sovereignty’ comes in. Kendall is clear that, for Britain, this does not mean sealing the borders to foreign tech or trying to build every part of the AI stack at home. Instead, she presents it as a plan to reduce dangerous dependence in the areas that matter most and to make Britain useful enough to others that it cannot be ignored. This is an important distinction for readers to notice. The speech is not selling isolation. It is selling influence through capability. Kendall says Britain should back more home-grown AI firms and work more closely with allies, especially other middle-power countries, on rules, standards and security. Put simply, the aim is not to do everything alone. It is to avoid being left with no room to choose.

The most concrete part of the speech is the industrial plan. Kendall says the Government’s Sovereign AI programme will put £500 million into British AI companies so they can start, grow and compete abroad. She also says selected firms will get fully funded access to the UK’s biggest supercomputers, faster routes for global research talent, support alongside the British Business Bank, and help through public procurement. For students, workers and founders, this is the practical argument hidden inside the slogan. Ministers are trying to use the state not only as a rule-setter, but as a customer, an investor and a gate-opener. The speech says the Ministry of Defence has ringfenced £400 million to support British innovation, including AI, because the Government wants Britain to be an AI maker rather than only an AI buyer.

Kendall also puts unusual stress on hardware. That matters because AI is often discussed as if it lives only in software, when it actually depends on physical machines: chips, memory, networking and energy. She says a new AI Hardware Plan will be launched at London Tech Week in June, and argues that Britain should not accept the idea that the chips race is already over. To make that case, the speech leans on British history and current firms. It points to the UK’s early role in computing, ARM’s processor design, and newer companies such as Fractile, Olix, Lumai, Optalysys and Salience Labs. Kendall also says ARIA is investing £100 million in scaling compute, including £50 million for an inference lab so start-ups can prove their hardware works. Her most eye-catching figure is that even a 5 per cent share of a future one-trillion-dollar AI chips market could bring in $50 billion. That is an ambition, not a guarantee, but it shows the size of the prize ministers think they are chasing.

The speech does not argue for British self-sufficiency in a vacuum. Kendall says Britain should work more closely with allies including Germany, France, Canada, Japan, Australia and South Korea, especially where countries can share research, investment and standards. She presents this as a route to resilience: if trusted partners build different strengths and co-ordinate them well, each country is less exposed when supply chains or security conditions change. **Why that matters:** ‘middle powers’ are countries that are influential, wealthy or scientifically strong, but are not in the same league as the United States or China. Kendall’s argument is that Britain can still matter a great deal if it becomes the partner others want in the rooms where rules, tools and security tests are being set. The speech is careful to say this should sit alongside, not instead of, the UK’s close relationship with the US.

Security is the other major thread running through the speech. Kendall describes the UK AI Security Institute as one of Britain’s strongest assets and says its role will grow as powerful models become better at finding cyber weaknesses. She points to Anthropic’s Mythos model, whose claims about discovering new cyber vulnerabilities were tested by the institute, and adds that other systems, including OpenAI’s GPT 5.5, have shown similar abilities. For readers, this is one of the clearest reminders that AI policy is not only about growth. It is also about inspection, testing and public trust. If models can write code faster, probe systems more effectively or copy expert behaviour at scale, then governments need ways to check what those models can actually do. Kendall says that in July the UK-chaired network of AI Security Institutes will publish best practice on the science of evaluating AI models. That may sound technical, but the question underneath it is simple: how do we test a system before it causes harm?

The closing message of the speech is political as much as technical. Kendall rejects calls to pause AI altogether and says the real choice is between shaping the technology or being shaped by it. In her version of events, stepping back would leave Britain poorer, less secure and more dependent on decisions made elsewhere. That is the case she wants the country to accept. But this is also where readers should stay alert. Big promises about national renewal always need public scrutiny. Who benefits from the investment? Which regions get the jobs? What checks protect workers, schools and public services from rushed deployment? The Common Room test is a simple one: if AI sovereignty is going to mean anything beyond a strong slogan, it has to give the public more say, more safety and a fairer share of the gains.