11 April 2026 · 3 min read · John
Using generic AI tools like ChatGPT on client work exposes your firm to hallucinated case law, data protection breaches, and professional sanctions. A purpose-built legal AI achieved 86% accuracy on UK legal questions in 2025 testing — ChatGPT 4.1 managed 54%. In law, a wrong answer can mean a wasted costs order, an SRA referral, or worse.
The case record is growing fast. In Ayinde v London Borough of Haringey, a barrister submitted five fabricated case citations produced by AI. The judge considered referring both the barrister and the instructing solicitor to their regulators. In Ndaryiyumvire v Birmingham City University, a wasted costs order was made after a firm filed pleadings citing two fictitious cases. By November 2025, the UK had recorded 24 incidents of AI-generated false citations — including the 20th employment tribunal hallucination case in a single month.
These are not stories about reckless lawyers. Most of them thought they were being efficient.
Then there is the data problem. One solicitor self-reported to the SRA after uploading client emails and Home Office decision letters to ChatGPT to summarise them. He told the tribunal he now understood this was a data breach. In June 2025, the Upper Tribunal confirmed in UK v Secretary of State for the Home Department that uploading privileged material to open-source AI tools waives legal professional privilege. That ruling applies to anyone doing the same thing tomorrow morning.
No. Your professional duties apply regardless of whether AI was used, and regardless of whether you used it personally or someone in your firm did. Your COLP is responsible for regulatory compliance when new technology is introduced. In a firm of ten or fifteen people, that is almost certainly you.
Only 9% of organisations feel confident their AI use is compliant. A third say they are not confident at all. And despite 31% of individual legal professionals using generative AI, only 21% of firms have any formal adoption policy. That gap is where the liability lives.
The Bar Council has been direct: AI-generated content that misleads the court — even accidentally — is gross negligence and could constitute contempt of court or perverting the course of justice.
I build AI tools configured specifically around your firm. Your precedents, your matter types, your tone, your workflows — not a general-purpose chatbot asked to guess what a UK solicitor needs.
Every tool includes proper data handling from the start. No client data going to consumer-tier AI accounts, no uncontrolled transfers, no guesswork on UK GDPR compliance. The Data (Use and Access) Act 2025 is already in force. ICO audits of AI systems are on the roadmap. Getting this right now is cheaper than fixing it after a complaint.
Small firms adopt legal-specific AI at roughly half the rate of larger ones. Partly resource constraints — but also because nobody has made it straightforward. That is what I do.
If you want to understand what AI could actually do in your firm without the risk, book a free 20-minute discovery call at book a call or email john@aigura.co.uk.
Want to see how this applies to your business?
Book a free 20-minute call →Solicitors can use AI tools but remain fully responsible for accuracy and data protection. ChatGPT 4.1 achieved only 54% accuracy on UK legal questions in 2025 testing. The Upper Tribunal confirmed in June 2025 that uploading privileged client documents to open-source AI waives legal professional privilege.
The SRA has not issued specific AI guidance, but its position is clear: professional duties apply regardless of whether AI was used. COLPs are responsible for regulatory compliance when new technology is introduced, and solicitors must ensure any information submitted to court is accurate and from verifiable sources.
Courts have issued wasted costs orders, referred lawyers to the SRA and Bar Standards Board, and the High Court has warned that submitting AI-generated false information could expose lawyers to criminal liability including contempt of court. By November 2025, the UK had recorded 24 incidents of AI-generated false citations.
Yes. Uploading client documents to consumer AI tools is a data protection breach under UK GDPR. One solicitor self-reported to the SRA after doing exactly this. In June 2025, the Upper Tribunal also confirmed that uploading privileged material to open-source AI tools waives legal professional privilege.