Artificial intelligence has quickly become part of everyday legal and tax practice. Solicitors increasingly use AI to review documents or assist with legal research. The issue now is how professional responsibility will be assessed when AI-assisted work proves defective.
The starting point is unlikely to surprise anyone. Courts and regulators are not creating a separate standard of care for advice or work product involving AI. Rather, established professional duties (competence, independent judgement, appropriate supervision, duties to the court and the obligation to act in the client’s best interests) are being applied in a new technological context.
The recent High Court decision in Cork & Anor v Smith [2026] EWHC 1199 (Ch) provides an early indication of how courts are likely to approach AI-related issues. The problem was not the use of AI itself, nor any inherent unreliability in the technology. Rather, material generated with AI assistance found its way before the court without being subjected to adequate verification. The firm concerned subsequently apologised and self-referred to the Solicitors Regulation Authority. The underlying point is a familiar one: responsibility for legal work remains with the solicitor. Regardless of how a document is produced, the solicitor who relies upon it or places it before the court remains accountable for its contents. AI may assist with drafting and analysis, but it does not displace professional judgement or responsibility.
For most practitioners, the greater risk is unlikely to be the type of error that attracts headlines. The profession has become familiar with stories involving fabricated authorities and fictitious legal propositions. Such mistakes are dramatic, but relatively easy to identify once uncovered. The more difficult issues are often far subtler. A clause appears legally sound but weakens the client’s position, or a summary captures the broad thrust of a document while omitting a qualification that materially affects the conclusion. These are precisely the sorts of defects capable of causing loss and, ultimately, giving rise to professional negligence claims.
One of the challenges with generative AI is that its output rarely looks problematic. Unlike poor legal work, which often reveals itself through clumsy drafting or obvious inconsistencies, AI-generated content is typically polished and coherent. That makes careful scrutiny more important than ever. The same is true of supervision. Work may appear well presented even where the underlying reasoning is incomplete, meaning supervisors may increasingly need to focus on what sources informed it and what steps were taken to verify the result. In some cases, the real question may not be whether AI made a mistake, but whether appropriate supervision would have identified it.
These themes featured prominently throughout London International Disputes Week 2026, in which the Law Society played an active role. One observation repeatedly made across several panels was that lawyers are increasingly being asked to validate analyses that have already been generated by large language models. Many clients now arrive having already consulted AI tools and formed preliminary views about a claim or dispute.
That development carries obvious liability implications. Once a solicitor endorses, adopts or relies upon an AI-generated analysis, responsibility for that advice rests with the solicitor. The fact that the analysis originated from a machine rather than a colleague, expert or client does not reduce the duty to verify it. Professional responsibility attaches to the advice ultimately given, not to the source from which the information was obtained.
LIDW also highlighted a broader consequence of AI adoption. The widespread use of AI-generated content is likely to significantly reduce the cost of producing legal work. As the cost of generating legal documents falls, the volume of legal activity is likely to increase. From a risk perspective, the use of AI therefore changes not only how legal services are delivered but also how mistakes can spread. An error that might once have affected a single matter can now be replicated across multiple matters within a short period. Greater efficiency brings considerable benefits, but also magnifies the potential consequences of defective processes.
The lesson emerging from Cork & Anor v Smith and the broader discussion across the profession is relatively straightforward. AI does not alter where responsibility lies. Solicitors remain responsible for the advice they give. What AI changes is the nature of the risks that require management and, potentially, the standard against which professional conduct is assessed. The negligence claim of the future may not arise because a model invented a case. It is more likely to arise because a key assumption went unchallenged or an apparently persuasive analysis was accepted without sufficient scrutiny. As AI becomes part of everyday practice, courts may increasingly ask whether a reasonably competent solicitor using such tools would have identified the problem before acting upon the output. Technology does not replace professional judgement but becomes one of the circumstances in which that judgement must be exercised.
None of this should be read as an argument against AI. Quite the opposite. Responsible adoption may itself become part of the expected standard of competent practice. As Sir Geoffrey Vos observed during LIDW, lawyers and courts are already struggling with the volume of information they must process, and technology will inevitably form part of the solution. Clients, meanwhile, are unlikely to pay for work that can be completed more efficiently through properly supervised technology. The challenge is therefore not whether AI should be used, but how it can be used responsibly without losing sight of the principles that underpin the profession. Sound judgement, integrity and a willingness to challenge assumptions have always been central to good legal practice. In the age of AI, they may become its defining features.
Cécile Perrault is a senior associate in the tax disputes team at Stewarts Law, London
No comments yet