UK Government Plans to Commercialize Public Data for AI Development

UK tech scheme includes AI tool to mark homework as ministers weigh selling data

An artificial intelligence tool to mark schoolchildren's homework is being developed using a trove of UK government data, ahead of plans to sell public records including health information within a decade. A new government scheme has pooled public documents such as curriculum guidance, lesson plans and anonymised pupil assessments, which are now being used to train AI models, including one that can mark children's work and provide detailed feedback. The "content store" -- created by UK company Faculty AI with £4mn of government investment -- serves as an early prototype for ministers as they draw up broader plans to sell anonymised public data to researchers and businesses within 10 years, science, Innovation and technology secretary Peter Kyle told the Financial Times. The program is "a really good example of the kind of gathering of data together and the use of data that . . . will ultimately be the National Data Library", he said. Sir Keir Starmer's Labour party set out plans in its manifesto to create a National Data Library which would aggregate and store troves of government data that could eventually be monetised. So far, companies, data experts and government officials have been unclear exactly what form this "library" will take, and whether it would simply be used to improve government record-keeping or if plans were afoot to sell the data to third parties. But Kyle said public data that could include anonymised health records "will be used for public benefit and part of that will be the commercialisation of it for scientific endeavour". "Some of it will lead to commercialisation and commercial activity, I'm not squeamish about that," he told the FT. "I will just be driven by making sure that it has public benefit." In January, the prime minister announced an "AI Opportunities Action Plan", which was written by Starmer's AI adviser Matt Clifford and recommended making five high-impact public data sets immediately available to the private sector and proactively collating data that would be useful to innovators. At the time, Starmer refused to answer questions about whether national data would be sold and how it would be priced. A report by the Tony Blair Institute think-tank about the National Data Library last month noted that "there is still no clear, bold vision for its purpose or how it should be delivered". Kyle suggested it would not be possible to set up the library within this parliament -- which is expected to run till 2029 -- but said he was "very confident" it could be done by the end of the "decade of national renewal" that Starmer pitched when he was elected. The government is currently midway through a six-month scoping exercise for the National Data Library, Kyle said. "The interaction between the private sector will happen only when we are able to reassure the public it will happen in a safe way and the benefits of doing so will be made absolutely clear to the public," Kyle said. The science secretary said that the data would remain in Britain and be used in the UK, though he was open to it being used by overseas researchers and companies. "What we will not have is companies from abroad coming in and using our data and taking it abroad," he said. Tom Nixon, managing director of applied AI at Faculty who was involved in developing the content store, said that "at the moment, we're giving access for free, but you can imagine over time you would have paid-for access to the data".

UK Public Records Could Be Commercialized Within a Decade, Says Science Secretary

The U.K. government wants to create a "National Data Library" of public records it can sell to AI developers. | Credit: Leon Neal / Getty Images. The U.K. government plans to create a "National Data Library" that could be used to monetize public records, including health and education data. Discussing the topic with the Financial Times, Technology Secretary Peter Kyle said he wasn't "squeamish" about commercializing public data, as long as there are civic benefits. Public Data Presents AI Opportunity Because civic records are generally subject to the highest privacy requirements, they represent one of the last untapped reservoirs of AI training data. But now, the U.K. government plans to open the country's public data sets to businesses and researchers building AI tools. In January, Kyle presented the AI Opportunities Action Plan to Parliament, calling for the public sector to collect more high-quality data and to "responsibly unlock" existing resources. More recently, the government launched a new program to equip data scientists working in the public sector with new AI skills. At the same time, efforts to boost public sector adoption of AI tools have become increasingly central to the government's ongoing efficiency drive. Models for the National Data Library The AI Action Plan points to examples of public databases from which the National Data Library could draw inspiration. These include the U.K. Biobank, a trove of biological samples, physical measurements, body and brain imaging data, bone density data, activity tracking and questionnaire answers collected from half a million people. In another example, Kyle pointed to a new Department for Education initiative to use student data to build AI tools for the classroom. Monetization on the Cards While there have been some efforts to organize public data for AI training in the past, the idea of charging for access is more novel. The U.K. Biobank provides a picture of what this might look like for health data. For three years of access to the platform, users pay up to £9,000, with discounts available for student researchers and customers from lower-income countries. Kyle acknowledged that similar anonymized health records "will be used for public benefit" as part of the proposed National Data Library. "Part of that will be the commercialization of it for scientific endeavor," he added. Although he didn't provide concrete details, the technology secretary suggested the government could start selling public data within a decade. Meanwhile, the AI action plan said the government should identify at least five "high-impact public datasets" it can rapidly make available. Working With British Institutions Alongside the nation's health and education systems, British cultural institutions could also provide a gold mine of unique data. For example, the government's action plan calls for a body like the BBC and British museums to contribute media assets that could be licensed to AI developers.