Digital Explainer – What is sovereign AI, and what is safe AI use in the public legal sector?

London Tech Week, it seemed, was all about what the UK Government will allow and what it sees as important and sovereign. The USA also made waves when its decision led to Anthropic, the developer of Claude, announcing it will “abruptly disable” its most advanced AI models for all users after the US Government ordered it to suspend access to the models for foreign nationals, citing national security concerns. Anthropic received the export control directive to suspend access to Fable 5 and Mythos 5 for all foreign nationals. Anthropic has suddenly become less accessible for the world.

In this Keynote, Technology partners James Tumbridge and Robert Peake give a wider understanding of what is concerning governments/the public sector when it comes to AI, and what they mean by sovereignty.

What is sovereignty?

AI sovereignty is an entity (business) or nation’s capacity to control AI technology it relies upon; this includes AI models, the data needed to train them, and the digital infrastructure connected with those AI tools. As global AI adoption increases, AI sovereignty has evolved from being principally about data residency, though that remains at the heart of many conversations. The AI systems now being adopted widely are integrated fully within organisations’ technology stacks, operating continuously; they often depend on sensitive data and proprietary models. As decision makers, you need to have a plan for business accountability, auditability, and data governance, as well as business continuity.

In the UK, digital sovereignty, is also called technology sovereignty, and it is defined as “the agency and capacity of any organisation to make intelligent, informed choices to shape its digital future by design”. This year, the Minister for AI defined sovereignty in the field of AI as “the ability for a state to have strategic leverage when it comes to this technology, such that it can ensure ongoing access to critical inputs, and ongoing assurance that its wider economic and national security objectives can be met more broadly”.

According to the UK Parliament, in the UK and Europe debates around digital sovereignty have arisen due to concerns around the potential security risks of Chinese technologies and, more recently, the market dominance of US ‘big tech’ companies. Presently, the UK Government does not have an overarching policy on digital sovereignty; it has set out its approach to building “sovereign capability” in key technologies in the June 2025 Modern Industrial Strategy and the accompanying Digital and Technologies Sector Plan.

What’s London saying?

Prime Minister Starmer framed Britain’s AI choices as a choice between three paths – head in the sand, no guardrails, or back British firms – and he claims the UK is backing the British, unveiling £400m for sovereign AI chips. Alongside this, BT, HSBC, Lloyds, NatWest, and BAE Systems are said to have signed up to help build a sovereign frontier model called Lumen.

Can AI be used safely in the public legal sector?

In a previous Keynote, we explained how UK judges are open to AI use. UK judges have Judicial AI Guidance, published in October 2025, with three key aspects:

• Judges are not prohibited from using AI.
• Judges must take full personal responsibility for whatever goes out in their name.
• A judge who wishes to use any kind of AI system should only use a system which they are sure is secure.

The Judicial Office, His Majesty’s Courts and Tribunals Service (HMCTS), and the Ministry of Justice’s AI unit have provided two AI systems which judges are satisfied are secure.

Judges hope to use AI for at least four purposes:

1. As a transcription tool.
2. To help produce anonymised judgments (a particular issue in family law, but not limited to that).
3. To identify internal inconsistencies in draft judgments during preparation.
4. For administrative tasks (using Copilot to find things in files more efficiently, for example).

However, use in criminal cases is now a concern, and the police have been told to be more cautious. The National Centre for AI in Policing, known as ‘Police.AI’, has said that police forces must stop using AI in the preparation of court statements, following concerns that improper inputs could contaminate legal proceedings. Police.AI issued the warning to ‘pause’ the use of AI after some forces were understood to have been using ‘commercially available’ AI tools for tasks such as creating witness statements from officers’ notes. Another use raising concern was the preparation of disclosure schedules listing evidence to be provided to the defence team in criminal proceedings.

A particularly worrying example of how AI can lead to errors occurred when the West Midlands Police last year relied on Microsoft’s Copilot in its planning around a football between Maccabi Tel Aviv and Aston Villa; the AI ‘hallucinated’ a past match involving Maccabi, which was included in the police dossier in support of banning the team’s fans from attending the planned match.

Alex Murray, Head of the Police.AI centre, said that police forces will now all have in place an AI policy that sets out the need to ‘check everything that it produces’. Regarding the opportunities afforded by AI in policing, he said: “I think the benefit that automation offers, with the appropriate guardrails, policy and training, outweighs the disadvantages.”

If you have questions or concerns about the use of AI, please contact James Tumbridge and Robert Peake.