AI

The Alan Turing Institute wants to build a UK AI model from scratch – and release it for free to the public

Ryan Brothwell 3 min read
The Alan Turing Institute wants to build a UK AI model from scratch – and release it for free to the public

Key Points

  • Alan Turing Institute proposes a free, open-source British AI model built from scratch
  • Government advisers want computing capacity set aside to trial it
  • CST warns the UK risks becoming just a high-value AI customer
  • Four recommendations include an AI Sovereignty Plan and a single accountable owner
  • Health and creative data named as areas to test data access

The Alan Turing Institute wants to build a British AI model from scratch and release it free to the public. The proposal features in advice from the Council for Science and Technology (CST), which the government published on Monday (19 October).

The CST described the model as national infrastructure, an open and UK-governed asset for researchers, start-ups, small businesses, public services and the open-source community. It added that the model would not be a proprietary product for a single firm or government department.

The council recommended that the government allocate some national AI computing capacity to trial different types of models and consider the Alan Turing Institute’s proposal.

The CST is the government’s top independent advisory body on science and technology, and it advises the Prime Minister and the Cabinet.

Its co-chairs, Dame Angela McLean and Lord Browne of Madingley, sent the advice to Cabinet Office minister Kanishka Narayan in a letter dated 30 July.

Warning over UK reliance on foreign AI

“Without deliberate and urgent action, the UK risks becoming a high-value AI customer, rather than a country able to shape the technology and its usage,” the council said.

The CST said the UK should keep access to leading global AI models but should not depend on them for every use case. It added that a specialised system developed in the UK would work better in some cases.

The council defined AI sovereignty as having enough control across the AI value chain to deliver the UK’s chosen outcomes, capture economic and public value, and stay resilient under pressure.

Key recommendations

The CST made four recommendations, starting with a UK AI Sovereignty Plan built on a map of the country’s AI ecosystem.

The map would cover computing power, data, skills, models, governance and public adoption, and identify which capabilities the UK should own, share or secure access to.

The council wants these choices to shape rapid AI pilots in areas such as national security, health and education.

Each pilot would need to answer six questions:

  • What problems need solving
  • What sovereign data the UK holds
  • What model capability it needs
  • What system the government must build around it
  • What value the UK can capture
  • How the government will measure success

The pilots would also test how different AI systems, including open-weight models and smaller domestic models, can handle different tasks.

The CST called on the government to appoint a single accountable owner for AI sovereignty, such as the head of the Cabinet Office AI Taskforce.

It also wants an AI skills and secondment route to sit alongside the government’s TechFirst programme, plus continued support for low-power computing research.

The council recommended that the government identify important UK datasets and test access arrangements in at least two areas, such as health and creative data.

The CST copied its letter to the Prime Minister, the Chancellor of the Exchequer and the secretaries of state for business, defence and culture.

Now read: Regulator wants ChatGPT to flag any time it gives money advice to Brits