On 16 June, technology minister Baroness Lloyd told the House of Lords that the government was investing £1.6 billion in Britain's "sovereign AI capabilities". The figure comprised £500 million for the Sovereign AI fund and £1.1 billion for an AI hardware plan.
Three weeks later, the House of Commons Science, Innovation and Technology Committee said ministers had still not set out which technological capabilities Britain needs to control, which it can share with allies and which it is prepared to obtain from overseas.
Its report, published on 7 July, found that the government had repeatedly described sovereign capability as a "critical priority" but had not explained precisely what that meant or how success would be measured. The committee said the 2025 industrial strategy had used the term without defining it and noted that other parliamentary committees had already raised the same question.
The government does have a broader definition.
Technology Secretary Liz Kendall said in April that British AI sovereignty did not mean attempting to build the entire technology supply chain domestically. She defined it as reducing excessive dependencies and increasing resilience in areas of national importance, giving Britain greater control and leverage while continuing to use foreign technology and attract overseas investment.
The government is not promising that Britain will produce all its own technology. On 1 June, Baroness Lloyd told peers that the £500 million Sovereign AI fund was not designed to replace foreign cloud providers. Its purpose, she said, was to reduce strategic dependence while helping British companies establish themselves in important parts of the AI industry.
Ministers have also acknowledged the scale of the existing dependence.
Answering a parliamentary question on 25 June, AI minister Kanishka Narayan said the UK currently relies on "a small number of foreign-headquartered providers for frontier models, compute and cloud". He said departments were responsible for managing the risks created by their own suppliers, with guidance from central government and the National Cyber Security Centre.
The £1.1 billion hardware plan is intended to build capacity further down that supply chain. It includes £750 million for a new national AI supercomputer, £120 million for an AI hardware innovation programme, new spending on skills and up to £150 million from the British Business Bank for a fund investing in hardware companies based in the UK. The government wants chips designed in Britain to form part of the new supercomputer, although the final suppliers have not yet been selected.
The argument over what Britain needs to control predates the latest report.
On 3 June, the same Commons committee published a separate investigation into digital government. It concluded that reliance on a small number of companies based in the US for digital infrastructure and public services was a strategic and economic vulnerability. It called for a definition of technology sovereignty shared across government and a list of the capabilities for which domestic control was considered necessary.
That report also recommended a technology sovereignty strategy setting out where the government wanted domestic alternatives to incumbent suppliers. It noted that developing wholly British alternatives would itself be difficult because UK technology companies can depend on American technology.
Events in the United States have since provided an unusually direct example of the dependency ministers are trying to manage.
On 12 June, the US government imposed export controls on Anthropic's new Fable 5 and Mythos 5 AI models, barring access by foreign nationals. Anthropic said the order took effect immediately and, because it could not reliably verify the nationality of every user in real time, it suspended both models for all customers.
The restriction was temporary. The US government lifted the controls on 30 June and Anthropic restored global access to Fable 5 from 1 July. Mythos 5, a less restricted version intended for specialist cybersecurity work, initially returned only to approved US organisations.
The Commons committee published its report a week later and still cited the episode as evidence of what could happen when access to an important technology ultimately depends on decisions made by another government. Its argument was not that the Anthropic models had been permanently lost to British users, but that access had been interrupted by an American decision with immediate effect.
The committee has recommended taking the old government framework of "own, collaborate, access" and applying it to individual parts of technology supply chains rather than labelling whole sectors as sovereign. That would mean specifying which components Britain intends to own, which it will develop with partners and which it expects to obtain from abroad.
As of 16 July, the government had announced the £500 million Sovereign AI fund and the £1.1 billion AI hardware plan, acknowledged its reliance on foreign providers and defined sovereignty in broad terms as reducing dependence and increasing leverage. It had not published the assessment, capability by capability, that Parliament is asking for: what Britain intends to own, where it will collaborate and where continued foreign access will be accepted.