“The SRA’s latest warning on artificial intelligence is a useful indication of just how quickly AI has moved from being a legal technology issue to a professional conduct issue.
The Law Society Gazette reports that the SRA received 42 reports of potential AI misuse between July 2025 and July 2026, with investigations concerning inaccurate legal citations, supervision and confidentiality. The SRA has now issued a formal warning notice. Its message is relatively straightforward: AI may be useful, but it does not dilute the solicitor’s personal and professional responsibility for the work ultimately produced.
The obvious example is the now familiar “hallucinated authority”. A generative AI system confidently produces a plausible case name, neutral citation and legal proposition, only for the authority not to exist. The problem is not confined to junior lawyers experimenting with new technology. The SRA says it has received reports from senior members of the judiciary and self-reports from solicitors, and recent cases have resulted in lawyers being referred to their regulators.
There is, however, a broader issue emerging in litigation which may prove at least as significant.
The third participant in the litigation
Solicitors increasingly need to think not simply about how their own firm uses AI, but how clients and opponents are using it.
A lay client can now upload documents, describe a grievance and within seconds receive a lengthy analysis of supposed causes of action, procedural objections and authorities. Some of those may be correct of course! A litigant in person can then produce a relatively professional formatted 20-page witness statement or skeleton argument at almost no financial cost.
That can undoubtedly improve access to justice. Indeed, the Lady Chief Justice has recently observed that judges sometimes find AI-assisted submissions from litigants in person clearer and easier to digest than submissions prepared without such assistance. The Master of the Rolls has gone further, predicting that AI will increase the volume of civil litigation because people who previously lacked the resources to turn a mass of information into an arguable claim can now do so using generative AI.
But there is another side to that development.
AI removes one of the traditional constraints on litigation: the time and cost involved in producing legal material. It is now extraordinarily easy to generate another letter, another argument, another application or another lengthy response. Where the output is inaccurate, repetitive or legally irrelevant, the burden of dealing with it does not disappear. It transfers to the opponent and ultimately to the court.
For solicitors acting against AI-assisted litigants in person, this can mean significant additional time being spent checking purported authorities, extracting genuine issues from lengthy documents and responding to points which have the appearance of sophisticated legal
arguments but little substance. The result, somewhat paradoxically, can be that a technology associated with reducing legal costs causes the costs of an individual dispute to increase.
The courts are plainly alive to the issue. Current judicial guidance expressly recognises that unrepresented litigants are using AI and may lack the ability to verify the legal information they receive. Judges are told that where AI appears to have been used they may ask what accuracy checks have been undertaken and remind the litigant that he or she remains responsible for the material placed before the court.
The Civil Justice Council is also examining AI in court documents and has identified litigants in person as presenting “distinct and evolving challenges”.
The regulators are converging
The Bar Standards Board’s May 2026 guidance adopts much the same basic philosophy as the SRA.
Barristers remain personally responsible for their work. AI-generated material should be independently verified. Particular care is required in relation to confidentiality, accuracy, court submissions and the use of general-purpose AI systems for sensitive material. The BSB regards court submissions and automated drafting as comparatively high-risk uses.
Perhaps more interestingly, the BSB devotes an entire section to “your responsibilities when others use AI”. Barristers are told to recognise that clients, instructing solicitors and opposing parties may be relying upon AI. The guidance specifically identifies opposing litigants in person who may use AI to advance legal arguments “without fully understanding their implications”.
That is an important development. AI competence is beginning to mean more than knowing how to operate an AI product safely. It increasingly includes knowing how to recognise and manage AI use elsewhere in the litigation.
What should solicitors do?
The answer is probably not to prohibit AI. That would be unrealistic and would sacrifice many of its genuine benefits. Instead, firms are adjusting ordinary litigation practices.
At the outset of a retainer, it may be sensible to establish whether a client has already been using AI to analyse the dispute or prepare material. Where a client arrives with a lengthy AI-generated “legal opinion”, the solicitor’s role is not simply to adopt or rebut it, but to establish which propositions are actually supported by the evidence and the law.
The same discipline should apply internally. Every authority relied upon should be verified from a proper source. Confidential client material should not be placed into an AI system without understanding precisely how the data will be processed and protected. The SRA’s warning makes clear that supervision is also critical: responsibility cannot simply be passed to the trainee, paralegal or AI product which produced the first draft.
There may also be a developing costs issue. Where an opponent produces large quantities of demonstrably unreliable AI-generated material, practitioners should keep a careful record of the additional work required to investigate and respond to it. Whether and when the courts will reflect such conduct in costs orders will inevitably develop with the case law.
The most useful way of viewing AI may therefore be neither as a threat nor as a substitute lawyer. It is becoming another participant in the litigation process. Sometimes it will make cases shorter, clearer and cheaper. Sometimes it will enable claims which would otherwise never have been brought. And sometimes it will generate ten pages of confident nonsense which somebody else has to spend two chargeable hours disproving.
For solicitors, the increasingly important skill will be knowing which of those three things has just landed in the inbox.”
Dilan Deeljur
September 2026
