The UK government has opened the first four competitions in a £100 million programme designed to buy and test British artificial intelligence technology for public services.
The initial challenges cover NHS productivity, more efficient computing infrastructure, the use of AI in defence environments, and security testing for AI agents, according to a government announcement published on Monday.
The money is available over the lifetime of the Sovereign AI Research and Development Procurement Scheme. It is not a £100 million award to one company or a commitment to spend the full amount on the first four challenges.
Official competition guidance says individual contracts can range from £250,000 to £10 million, although the government expects most to be worth between £1 million and £3 million. The first application deadline is 1 October 2026, with two further assessment rounds due to close on 1 December 2026 and 1 February 2027.
Government becomes an early customer
The scheme is a form of research and development procurement. Rather than giving companies a conventional grant, the government will contract them to design, test and demonstrate technology that addresses a defined public-sector problem.
Projects must devote at least half of their contract value directly to R&D services and are aimed at technology readiness levels four to eight. In practical terms, this means the technology should have moved beyond basic research but may still require testing before full deployment.
The Sovereign AI scheme page says applicants must be UK-registered start-ups or small and medium-sized businesses developing AI capabilities in the UK. There is no minimum turnover, trading history, net asset or cash reserve requirement.
Upfront payments may be offered where a cash-constrained company can justify them. Suppliers will retain the intellectual property created during a project, while the government will receive a licence to use the outputs, data and results for public-sector purposes.
Those terms address two obstacles that often make government contracts difficult for young companies: financing work before being paid and giving up control of technology that could later be sold elsewhere.
Four public-sector tests
The NHS competition, run with the Department of Health and Social Care, is seeking systems that can automate workflows, coordinate care or support decision-making. A separate challenge, involving the Department for Business, Innovation, Science and Trade and ARIA’s Scaling Inference Lab, is focused on reducing the cost and improving the efficiency of public AI computing infrastructure.
The Ministry of Defence wants open, secure and modular AI capabilities that can connect data and operate across defence systems. The National Cyber Security Centre is supporting the fourth challenge, which seeks methods for assessing and reducing the security and resilience risks created by increasingly capable AI agents.
An AI agent is software that can plan and carry out a sequence of tasks with a degree of autonomy. That capability can be useful, but it also creates risks if the system takes an unintended action, handles sensitive data poorly or is manipulated by an attacker.
Companies must first join an approved supplier list before submitting a full proposal. Applications will then be assessed by the Sovereign AI team, the department responsible for the relevant challenge and independent technical experts.
The government said the first competitions would test the procurement model before further challenges were added. Winning a contract would fund development and provide a public-sector customer, but it would not guarantee that the technology is later deployed across government.
That distinction is important. The scheme may help smaller firms prove that their products work in operational settings, while the larger economic benefit will depend on whether those demonstrations lead to reliable services and commercially viable products.
