In summary

AI can support litigation, but lawyers are responsible for accuracy and professional judgement. For the SRA, an AI error may expose failures of competence, supervision and governance. Managers, COLPs and supervisors must be able to evidence approved training, tools and oversight to verify information.

Properly deployed, AI has the potential to improve efficiency and enhance client service. However, the Divisional Court’s decision in R (Ayinde) v London Borough of Haringey; Al-Haroun v Qatar National Bank QPSC [2025] EWHC 1383 (Admin) is a strong reminder that technological innovation does not change a lawyer’s fundamental professional obligations.

For solicitors, law firms and managers, this is not just another case about fabricated authorities. It is a significant decision about governance and accountability. It is likely to become an important point of reference when regulators examine misconduct arising from the misuse of AI.

More than a 'fake case' story

The Divisional Court considered two separate matters: the court’s inherent jurisdiction to regulate proceedings and enforce lawyers’ duties to the court. Material presented to the court contained legal references that were either fictitious or unreliable, giving rise to concerns about the use of generative AI and the adequacy of subsequent checks.

The Court recognised the growing importance of AI within legal practice and was careful not to condemn the technology. Dame Victoria Sharp P and Johnson J observed:

“Artificial intelligence is a powerful technology. It can be a useful tool in litigation, both civil and criminal.”

The statement is important. The Court did not suggest lawyers should avoid AI altogether. Rather, it addressed the professional consequences of relying upon AI-generated material without adequate verification.

The Court identified concerns extending beyond the conduct of the individual practitioners involved:

“The facts of these cases raise concerns about the competence and conduct of the individual lawyers who have been referred to this court. They raise broader areas of concern however, as to the adequacy of the training, supervision and regulation of those who practise before the courts.”

That passage may ultimately prove to be the most important part of the judgment from a regulatory perspective.

AI as the digital equivalent of an unsupervised paralegal

The easiest way to understand Ayinde is to think of generative AI as an exceptionally capable but entirely unsupervised paralegal. It can draft witness statements, generate legal arguments, prepare chronologies, summarise hundreds of pages of documentation, and appear to undertake legal research.

However, it can also fabricate citations, misquote authorities, misunderstand legal principles, and express complete confidence in entirely incorrect conclusions. It will try its best to give you an answer and won't always qualify its response. It does not necessarily consider all of the pages in a large bundle of documents. It may not be able to open attachments or interrogate data to the required depth, and it probably won't tell you this unless you specifically ask it to explain what it was (not) able to achieve.

Competent solicitors do not file a pleading drafted by a trainee or paralegal without checking its contents, yet many practitioners appear willing to place substantial trust in AI-generated material.

The Divisional Court’s response was unequivocal:

“Those who use artificial intelligence to conduct legal research, notwithstanding these risks, have a professional duty therefore to check the accuracy of such research by reference to authoritative sources, before using it in the course of their professional work.”

The duty applies in advance of providing advice to clients and before any material is placed before a court.

Why the judgment matters to the SRA

From an SRA enforcement perspective, the significance extends far beyond the specific errors. The most obvious regulatory issue is competence, but the case potentially engages a much wider range of regulatory obligations. Central to this is integrity and the maintenance of public trust and confidence. As the SRA develops its strategy, it will also focus on systems and managerial oversight.

The SRA’s warning notice on the use of artificial intelligence adopts the same approach. It does not suggest that AI should be avoided. Instead, it underlines that regulated individuals and firms remain accountable for compliance with professional obligations when using technology.

Accordingly, an AI-related incident is unlikely to be viewed purely as an isolated error committed by an individual fee earner. Regulators are increasingly likely to ask broader questions:

1. What training was provided?
2. Which AI tools were approved?
3. Who was responsible for supervision?
4. What verification processes existed?
5. How were AI risks assessed?

The answers may determine whether a matter is viewed as an individual mistake or a wider compliance failure.

The growing importance of supervision

Perhaps the most significant lesson emerging from Ayinde is the renewed emphasis on supervision. Historically, investigations often focused primarily upon the individual solicitor responsible for a particular document or course of conduct. AI changes that dynamic. Wider-ranging deployment of AI is likely to increase the number of junior lawyers a senior lawyer can supervise.

Large numbers of fee earners may be using identical tools and similar prompts across an entire department. The resulting risks can therefore become systemic rather than individual. The Divisional Court’s focus upon “training, supervision and regulation” should be regarded as a warning to managing partners, heads of department and compliance officers.

Future regulatory investigations may increasingly focus upon what managers knew, what safeguards were implemented and whether firms exercised proper oversight in the deployment of AI technology.

The connection with Mazur and the conduct of litigation

The decision sits alongside the courts’ recent focus on personal accountability for the conduct of litigation. Although arising in a different context, Mazur v Charles Russell Speechlys LLP [2025] EWHC 2341 (KB) was concerned with who may conduct litigation and the limits of delegation to non-admitted employees.

The common practical theme is responsibility. Delegation does not eliminate the obligations attached to authorised practice. Whether work is performed by a trainee solicitor, a paralegal, an agent, an external provider or an artificial intelligence system, the authorised professional and the regulated firm must understand who is doing what, under whose authority and with what supervision.

Courts and regulators are unlikely to be persuaded by explanations that merely shift responsibility to a junior colleague, a client or a piece of technology.

Lessons for trainees, paralegals and junior lawyers

Junior lawyers should not regard Ayinde as a rejection of technology. The Court acknowledged that AI can be a very useful tool. The real point is that AI output must be treated as a draft rather than a finished product.

• Authorities should be located in an authoritative source.
• Quotations should be checked against the original text.
• Legal propositions should be tested by conventional legal research.
• Summaries should be cross-referenced with the underlying material.
• Uncertainty should be escalated to the supervising solicitor before the work is deployed.

Used in this way, AI can enhance efficiency while maintaining professional standards. Used unquestioningly, it may create substantial regulatory risk.

Implications beyond the legal profession

The principles discussed in Ayinde are unlikely to remain confined to SRA regulation. Similar issues will arise wherever regulated professionals use AI-generated work.

The underlying regulatory proposition is simple. Technology does not replace professional judgement or dilute personal accountability. Other regulators are therefore likely to ask many of the same questions about competence, assurance, record keeping, confidentiality, candour and supervision when AI contributes to a professional failure.

Could AI misuse lead to contempt proceedings?

Although the Court ultimately declined to initiate contempt proceedings in these cases, practitioners should not assume that future cases will be treated in the same way. The Court was very critical of the failures identified and voiced concern about the wider consequences:

“There are serious implications for the administration of justice and public confidence in the justice system if artificial intelligence is misused.”

The issue is also relevant to the CPS and other prosecuting authorities considering potential contempt. Ayinde demonstrates a need for careful analysis of the evidence and the applicable threshold. An inaccurate AI-generated submission may involve incompetence, a professional conduct breach, contempt, or some combination. The categories should not be conflated. The evidential basis for knowledge, intention or recklessness will require particularly close attention.

What firms should be doing now

The Law Society’s guidance, Generative AI: the essentials, takes a balanced approach. It recognises opportunities for the profession while identifying legal, ethical, regulatory and practical risks. Its guidance reinforces the importance of understanding the tools, protecting confidential information, checking outputs and implementing proper governance. Firms should therefore ensure that they have:

• a clear AI policy identifying approved and prohibited uses;
• an approval and procurement process for AI tools;
• role-appropriate training for partners, supervisors and junior staff;
• verification requirements for cases, quotations and legal propositions;
• confidentiality, privilege, data protection and information security safeguards;
• supervision arrangements that properly reflect the risk and experience of the user;
• an audit trail capable of showing how important outputs were checked; and
• a documented response plan for errors, data incidents and potential reporting obligations.

The objective is not to create bureaucracy for its own sake. It is to ensure that the firm can explain, to a court, client, insurer or regulator, why the use was appropriate and how the risks were controlled.

Conclusion

Ayinde v Haringey is not fundamentally a case about technology. It’s really a case about professional responsibility. Artificial intelligence can assist lawyers, but it cannot owe duties to clients, or answer regulatory allegations. Most importantly, it cannot hold a practising certificate.

For managers, supervisors and compliance officers, the judgment is a warning that AI governance is becoming a regulatory issue in its own right. For trainees and junior lawyers, it is a reminder that technology may help, but it cannot replace professional scepticism and independent verification.

As AI becomes increasingly embedded within legal practice, the distinction may prove to be one of the most important professional conduct lessons of the decade.

How JMW can help

JMW’s professional regulation team advises solicitors, law firms, managers, COLPs and other regulated professionals facing regulatory investigations and disciplinary proceedings. We can assist with responses to a regulator’s enquiries, self-reporting issues, court or tribunal referrals, AI-related confidentiality incidents, remediation, governance reviews and representations concerning competence or supervision.

Early advice can help identify the relevant duties, preserve evidence, assess any reporting obligation and present a measured response supported by credible remedial action.

Evan Wright is a JMW partner, barrister, mediator and Law Society Lexcel accredited consultant.