--- title: "UK’s Legal AI Sandbox Closes Applications as 83% of Lawyers Flag Hallucination Risk" date: 2026-10-06 author: "Sofia Ramirez" featured_image: "https://sqmagazine.co.uk/wp-content/uploads/2026/10/uk-legal-ai-sandbox-hallucination-risk-featured.jpg" categories: - name: "Artificial Intelligence" url: "/artificial-intelligence.md" tags: - name: "SP" url: "/tag/sp.md" --- # UK’s Legal AI Sandbox Closes Applications as 83% of Lawyers Flag Hallucination Risk On 27 September this year, [applications for the UK government’s advisory AI Growth Lab closed](https://www.gov.uk/government/publications/advisory-ai-growth-lab-legal-services/legal-services-advisory-ai-growth-lab-overview). The lab was designed as a regulatory sandbox that will first attempt to tackle issues currently plaguing the legal sector. Applicants who have placed bids already can expect to hear the results of their applications by the end of October. The scheme could not have been set up at a better time. AI is permeating all aspects of law, and many companies and clients are using it to check their legal standing and provide more efficient services. A strange dynamic is underway. Just as the use of AI by lawyers is becoming universal, fears around hallucinations are also reaching an all-time high. Most legal professionals believe that AI hallucinates in a legal context and are worried that these problems with the software will lead to damaging outcomes for their practices and their clients. ![UK legal AI sandbox and AI hallucination risk for lawyers](https://sqmagazine.co.uk/wp-content/uploads/2026/10/uk-legal-ai-sandbox-growth-lab.jpg)[Unsplash – CC0 License](https://unsplash.com/photos/a-persons-head-with-a-circuit-board-in-front-of-it-WhAQMsdRKMI) ## What the Growth Lab Will Bring The Growth Lab is designed to bring multiple public bodies to a single table. It will work with: - the Department for Business, Innovation, Science and Trade - the Information Commissioner’s Office - the Ministry of Justice - the Legal Services Board - the Council for Licensed Conveyancers - the Solicitors Regulation Authority It will serve as a single point of contact for companies building AI products for legal work. Rather than having to deal with multiple regulators at the same time, the sandbox will help AI labs put forward their solutions to a single body. Grapple Law, which offers assistance with employment law, said the following: > “Previously, if a company wanted to develop a [legal AI tool](https://www.grapple.law/), they had to go through a gruelling approval process. The Growth Lab aims to change that. While it’s in sandbox form at the moment, there will hopefully come a point where it becomes policy, streamlining the process while keeping both clients and legal firms safe. Tech companies won’t need to deal with several regulators at once, each with their own rules. Instead, they will be able to test their products in an enclosed environment and face the scrutiny of multiple bodies before approval.” The UK government is currently listing various types of projects that it wants entrepreneurs to take on. For example, it’s looking for solutions that help lawyers work more efficiently and get more tasks done. It also wants software that can effectively analyse and manage legal information and improve people’s access to justice, which can be limited in some situations. The stated aim is to make services better, faster, and more affordable for members of the public, but companies taking part won’t necessarily get any regulatory approval or endorsements. Ultimately, the lab is only offering advice at this stage rather than a licence. ## The Numbers That Matter The reason for the founding of the Growth Lab is simple. According to [the Scaling Up report published by LexisNexis](https://www.legalcheek.com/2026/09/hallucinations-top-lawyers-ai-worry-list-research-finds/), **94%** of lawyers now use AI for legal work. Around 75% use it at least once a week, and a third say that they log on to applications like ChatGPT and Claude every day. It’s not just for legal research, either. While this is the most common use case, around 62% of legal professionals say that they also use AI for document summarisation. Around half also use it for drafting and document review. The problem for many legal professionals is accuracy. Around **83%** of lawyers say that they are concerned about AI hallucinating or fabricating information that they then pass on to clients or courts. Interestingly, at the start of 2024, that figure was only 57%. Since then, lawyers have realised that AI can be highly prone to making mistakes, especially on long-form tasks. Some are worried about confidential data leaks, while others are concerned that they are relying too much on technology and failing to provide traditional legal practice services to their clients. At the same time, comfort with using AI is also increasing. Around **81%** of respondents to the LexisNexis survey said that they feel more comfortable with AI that’s grounded in legal sources. That’s up from 70% a year earlier, showing that specialist legal-specific artificial intelligence technologies can provide improved performance and utility. ## The Client Is Facing Issues AI is also causing various client-facing issues, which is another reason why the Growth Lab’s appearance on the scene is so important. Around three-quarters of legal professionals think that AI is causing clients and members of the public to misinterpret legal statutes. While artificial intelligence can be a good assistant for providing more information on legal standing, it doesn’t necessarily understand every step of the law perfectly. Sometimes it can create issues of its own. It may provide clients with more confidence than is warranted or brush off a legal issue that actually has significant weight. Even so, [AI is a significant part of life in the UK](https://sqmagazine.co.uk/ai-in-everyday-life-in-the-uk/) right now, despite only becoming viable three or four years ago. It’s used for bank chatbots and NHS triage, as well as providing assistance with insurance claims and tenancy agreements. The problem is that many AIs originated outside of Britain, and their text and responses were not written with English law in mind. In many cases, answers can sound confident but can also include legal terms from the US or other countries. This is specifically where product design matters most. Solicitors in the UK are keen on AI as long as it’s trained on British law and statutes. They’re not so happy with generalised models that have knowledge of legal systems from across the world, including in countries like China, Japan, and Turkmenistan. ## Who is liable? Another issue currently hanging over the entire sector is who is liable for AI-assisted or AI-related harm. The UK Jurisdiction Taskforce took on this question in its [legal statement on liability for AI harms](https://www.scl.org/ukjt-says-english-law-already-has-the-tools-to-address-ai-harms/). It concluded that most English law can already deal with AI-related harm, and that AI has no legal personality of its own. For professional lawyers, this means they need to continue using AI carefully. If they use it carelessly and fail to check its outputs, they could be held liable for harming their clients.