The UK is beginning 2026 with big artificial intelligence (AI) ambitions, but without a clear approach to the data that underpins them. Despite some recent positive statements of intent from the government, data needs more attention, particularly to make it AI-ready.
The last year has seen consultations and strategies galore, including the AI Opportunities Action Plan, the 10-Year Health Plan for England, the Modern Industrial Strategy and the Medicines and Healthcare products Regulatory Agency’s (MHRA) consultation on the Regulation of AI in Healthcare.
Many of these are either directly related to AI or depend heavily on it to achieve their aims. Yet the data on which their goal achievement relies is often inconsistent, incomplete, or hard to use at scale.
The latest – welcome – addition this year is the Roadmap for Modern Digital Government, which translates the earlier Blueprint for Modern Digital Government’s commitment to ‘completely reshape public services through digital transformation’ into specific departmental commitments and timelines running to 2030.
The UK has strong data foundations, including a long-standing open data tradition and some of the most valuable public datasets anywhere in the world, including health records, longitudinal education data, official statistics, and cultural heritage data.
The last Budget focused on growth, tax modernisation and digital services with the delivery of almost every measure linked to better data use – but much of the data that the government expects AI to rely on is not ready for that purpose.
The ODI’s 2025 report, the UK government as a data provider for AI, revealed that large volumes of government-published data cannot be reliably accessed by large-scale AI systems such as CommonCrawl, which is used by many major AI platforms, including ChatGPT and Anthropic.
This could lead to incomplete or misleading outputs if those systems are used to access, for example, benefits advice and health information. It could ultimately negatively impact people’s lives if they receive inaccurate information, because official sources cannot be read by AI.
At the local level, this could also affect day-to-day service delivery. Local authorities generate – and use – vast amounts of data in the delivery of social care, housing, transport and community services.
There are opportunities for hard-pressed councils to make financial savings by using that data as the raw material for automation, forecasting and citizen-facing digital tools. However, our recent report, Insights from UK councils on standards, readiness and reform to modernise public data for AI, published jointly with Nortal, found that many councils are still publishing data in formats designed for reporting rather than advanced technical use, which severely restricts the ability to use AI and digital tools.
The challenges range from inconsistent formatting and limited metadata to insufficient infrastructure, including missing Application Programming Interfaces (APIs), search functionality, and version control.
Common failings came through in the work, but there is also an emerging roadmap for progress. Our report suggests that “by modernising metadata, adopting standards and enabling secure data sharing, councils can move from ‘AI in theory’ to ‘AI that delivers on its promise’ – services that are efficient, transparent and trusted.” Indeed, several councils are working to improve data quality, strengthen infrastructure, and experiment with new approaches.
Building on these developments will be key to unlocking AI’s potential and ensuring public datasets contribute more fully to operational efficiency, service quality and policy effectiveness.
The establishment of GDS Local in November 2025 to connect central and local government, and the Ministry of Housing, Communities and Local Government (MHCLG)’s completion of a discovery into GOV.UK One Login for local government (working with 50 local authorities), are positive steps, but the structural challenges of inconsistent data, limited infrastructure, and supplier concentration remain significant.
Earlier this year, the Department for Science, Innovation and Technology (DSIT) published its Guidelines and best practices for making government datasets AI-ready, setting out high-level principles, success criteria, and examples of excellence to ensure that public sector datasets are AI-ready.
The guidelines cite models such as the ODI’s Framework for AI-Ready Data, which recommends concrete steps to help organisations optimise data for machine learning by improving its overall quality and adherence to standards, and ensuring its legal compliance and responsible collection.
Other key questions about the public sector’s data ecosystem need to be answered in the year ahead.
For example, despite £100 million in funding announced for the National Data Library in June 2025, and a commitment to share the programme’s vision and ‘kickstarter projects’ in early 2026, key questions are unresolved, including whether it will operate as public infrastructure or adopt a commercial model in which certain categories of users will pay to access the data it contains. This fundamental choice will shape access, licensing, and accountability.
The Data (Use and Access) Act sets out frameworks for smart data and digital verification services, but the Government still faces complex issues around AI and copyright.
So there’s a potential paradox emerging in which public data is treated alternately as a public good and a monetisable asset: “sovereign” in theory but undefined in practice, with technical innovation moving faster than the guardrails needed to ensure quality, share value, and protect our national interests and individual privacy.
This is clearest in health and social mobility data. The NHS’s health data is a priceless asset, yet much of its value has not been realised, partly because of people’s understandable concerns around privacy.
But there are ways that other kinds of – non-personal – data can help improve health outcomes and support the three “shifts” set out in the 10 Year Health Plan (hospital to community, analogue to digital, and sickness to prevention).
OpenActive, for example, enables millions of physical activity sessions to be published every month. This data could be embedded in clinical pathways or social prescriptions and could power services that help people to get active, and be made available through the NHS App. Physical inactivity has been calculated to cost the UK over £20 billion a year, so data initiatives like OpenActive have the potential to both improve health outcomes and save money.
In social mobility, rich public datasets such as the Longitudinal Educational Outcomes dataset, the Parent Pupil Match, and the PAYE dataset exist but are not coherently linked, leaving policymakers unable to answer basic questions, such as why young people become NEET or how income and geography shape opportunities for children.
The problem is not a lack of data, but a lack of joining up around how this data should be shared and used for public benefit.
Many of the challenges are acknowledged in September’s Rewiring the state: Unlocking government transformation study. The Roadmap for Digital Government answers them in part by setting out specific ambitions and commitments, including the Department for Work and Pension’s (DWP) plan to modernise priority legacy benefits services, DEFRA’s migration of 19 critical applications to the cloud by March 2026, and improved asset management standards this year.
But questions remain about whether the pace and funding will match the scale of the challenge. The Roadmap for Digital Government undoubtedly moves us closer, but there is more to do.
Data is increasingly framed as foundational national infrastructure alongside power grids and transport networks. The Roadmap commits to defining core data standards for interoperability and AI, and mandating API adoption across the public sector by March 2026, but these foundations will only succeed with sustained investment in assurance, stewardship, and the skills to put them into practice.
At the moment, the UK underinvests in these quieter but essential capabilities, and as a result, its datasets are not yet capable of delivering the social and economic value that they could.
If 2026 is to be a year when digital and data ambitions are realised, decisions are needed about which public datasets are national infrastructure, which should be treated as public goods, and how value can be realised – and shared – without eroding public trust.
This also means investing in the essential work of implementing data standards, governance, and digital and data skills to ensure AI systems are built on solid foundations.
The Roadmap’s target of one in ten civil servants working in digital, data and cyber roles within five years, and a core digital curriculum for all civil servants by April 2026, are welcome, but will require sustained commitment to ensure AI systems are built on solid foundations.
The UK has the data assets and the expertise; what it needs to do now is decide – deliberately and transparently – what kind of data nation it intends to be.