On 17 August 2026 the Solicitors Regulation Authority published a warning notice on the misuse of AI. Three days later, on 20 August, it published its fourth annual assessment of continuing competence. The second document does not mention artificial intelligence anywhere. I searched the full text. The words AI, artificial intelligence and technology do not appear in it. The word digital appears once, describing the channels through which the SRA signposts its own resources. That is the whole of it.
I want to be careful about what I am claiming. This is not an accusation of inconsistency, and it is certainly not a suggestion that the SRA has done anything wrong. The point is narrower and, I think, more interesting: the machinery we use to assure competence is still looking at the profession as it was, while the same regulator has just described a risk that lives inside how lawyers now actually work.
There is a second reading, and I want to be careful with it too. The Serious Event Review into the SRA’s handling of the PM Law Group, delivered by Jenner & Block and published on 3 September 2026, found that the regulator held more information about the group than was ever drawn together into a single coherent picture, and that this reflected not individual error but “a structural limitation in the SRA’s operational architecture”. That finding is about supervision, not about competence policy, and I am not going to pretend it reaches across to these two documents. But a regulator that has just been told it struggles to connect what it holds in one part of the organisation is worth watching for the same difficulty elsewhere. Two documents published three days apart, one describing a risk and the other examining how solicitors keep themselves competent, is a place to watch.
What the warning notice does
Warning notices are not discussion papers. The AI notice’s status section states that it is designed to help solicitors and firms understand their obligations and how to comply with them, and that the SRA “will have regard to it when exercising our regulatory functions.”
It raises two concerns. The first is hallucination: false content and false citations reaching the court. The SRA records that it has “received reports of potential breaches of our Code of Conduct from senior members of the judiciary”, and that solicitors have also self reported having relied on tools that produced inaccurate or misleading content. The second is confidentiality, where client information is entered into an AI tool.
Then comes the sentence that decides everything else in the document. “AI has no separate legal personality; solicitors and regulated individuals who use AI in the course of delivering legal services remain accountable for their work and outputs, regardless of how that work has been prepared.”
Notice where the SRA then goes looking for the obligations that carry that accountability. Not to a new technology rulebook. To paragraph 3.2 of the Code, on providing a competent service; to paragraph 3.5, under which those who supervise or manage others remain accountable for the work carried out through them and must effectively supervise it; and to paragraph 3.6, which requires solicitors to ensure that the individuals they manage are competent to carry out their role and keep their professional knowledge and skills, and their understanding of their legal, ethical and regulatory obligations, up to date.
The notice is blunt about what follows for supervisors. Those who supervise junior or non authorised colleagues “may also be found to have breached regulatory requirements and professional duties if false citations are put before the court without adequate review and/or supervision.” The regulator’s own summary of the position is that “appropriate human oversight, informed professional judgement and a proportionate, risk-based approach are essential to ensure compliance with your regulatory and legal obligations.”
So the SRA has located AI risk squarely inside the competence and supervision provisions of the Code. That is a deliberate and, in my view, a correct choice. It is also a choice with consequences, because it means the question of whether a solicitor is handling AI properly is now a continuing competence question.
What the annual assessment covers
The assessment published three days later is a serious document and it is candid in places where it did not have to be. Competence related reports rose from 2,720 in 2024 to 4,038 in 2025, though they stayed stable at 23 per cent of all reports because the overall number rose too. The SRA reviewed 123 training records from solicitors practising criminal and civil law across 66 firms, covering 1 October 2024 to 30 September 2025. All 123 had done some learning and development. Almost all had kept a record.
Two findings deserve more attention than they have had. Most solicitors who kept a record had completed between one and four learning activities, with the focus on maintaining technical legal practice; and, in the SRA’s own words, “our review did not assess the quality or effectiveness of the learning and development carried out.” Just under half of the records reviewed did not explain how the learning need was identified and why it was important to address.
Now the fair objection, which I will make against myself. The report analyses reports received between 1 January 2024 and 31 December 2025. AI issues at scale may simply sit outside that window, and a regulator cannot report on what its data does not yet show.
I do not think the objection survives contact with the document. The assessment does not draw only on that data. It also draws, in its own description of its sources, on information from the SRA’s wider regulatory activities “including our horizon scanning programme to explore emerging and future developments in the legal services market”. It carries a section headed what the SRA will do next. It describes forward looking work on ethics intended, among other things, to “reflect the realities of modern legal practice”. None of those sections mentions AI either. The silence is not confined to the parts of the report that look backwards.
The Bar has already written the connection down
The comparison worth making here is not solicitors against barristers on performance. It is one regulator’s architecture against another’s.
The BSB’s Guidance on the use of Artificial Intelligence and Other Technologies, valid from 18 May 2026, states that under Core Duty 7 “you should maintain a sufficient level of competence in technology and AI to understand how they may impact your practice, whether or not you adopt those technologies yourself.” That expectation is not left floating. The guidance says in terms that it derives from “Core Duty 7 (Competent standard of work), Core Duty 10 (Practice management), ongoing competence expectations and CPD obligations”, and it names the competence engaged as the Professional Statement for Barristers, competence 1.12, employ effective research skills, and within it sub point (a), accurately identify the legal and non-legal issues.
That is the difference, and it is a difference of design. The BSB has run technology and AI competence into its continuing competence and CPD expectations, so it arrives in a barrister’s learning plan by the ordinary route. The SRA has run it into a warning notice that it will have regard to when exercising its regulatory functions, while the document that examines how solicitors maintain their competence says nothing about it. One regulator has put the requirement where lawyers plan their learning. The other has put it where lawyers are held to account. I am not going to pretend I know which will work better. But they are not the same thing, and only one of them shapes behaviour before something goes wrong.
Why this matters now rather than in 2028
Subject to the analysis of consultation feedback and to approval by the Legal Services Board, the SRA will require all solicitors to keep a record of their learning and development from the start of the 2027/28 practising year. The rule as consulted on requires the record to show that the solicitor has considered their role and work, identified any learning needs, and recorded how those needs were addressed. The word reflection does not appear in either the proposed rule or the record content consulted on. That is odd, because the SRA’s own resources explain “what reflection looks like and when and how to do it so that all learning and development needs are identified”, it provides a template solicitors can use to “capture their reflection”, and the first of the four failings it lists below is an inability to show that learning and development came from regular and meaningful reflection.
That rule is being finalised now. What it requires will decide what hundreds of thousands of records contain for years afterwards, and records are sticky: once a format is in the field, it changes slowly.
The SRA lists four things it consistently sees in some solicitors. They cannot demonstrate that their learning and development results from regular and meaningful reflection. They focus on maintaining technical legal knowledge rather than the wider skills competence requires. They have not demonstrated regular learning and development relating to ethics. And they “are not aware, or have not taken account of, relevant warning notices and guidance when identifying their learning needs.”
Read that fourth one again with the date in mind. A warning notice on the misuse of AI was published on 17 August. The report identifying failure to take account of warning notices as a persistent weakness was published on 20 August. The profession has just been handed an unusually clean worked example of the very failing the regulator says it keeps seeing, and the report that names the failing does not connect the two.
The question I cannot answer
The SRA has left itself room here, and it says so openly: “this is our fourth annual assessment of competence. We are currently reviewing how we develop this. We may change our approach to, and format of, this report going forward depending on the outcome of this work.” This is a door the regulator has already said is open.
So here is what I keep getting stuck on. Suppose a solicitor supervises a junior who uses AI to prepare a first draft. Under paragraph 3.5 that solicitor remains accountable for the work. Under the warning notice, appropriate human oversight is essential. What would the file record and the junior and supervisor’s learning records have to contain before the SRA, or a firm, or the practitioners themselves, could tell whether the oversight is any good?
An attendance line saying they went to a session on AI does not answer it. Nor, I suspect, does a reflective paragraph written twelve months later. Something recorded much closer to the moment might: a short note of what was accepted, what was changed, what was rejected and why, written while the decision was still fresh. That is roughly what paragraph 7.2 of the Code of Conduct for Solicitors, RELs, RFLs and RSLs already expects of a solicitor in any event, since it requires them to be able to justify their decisions and actions.
I have views, and I am not yet confident enough in them to publish them as an answer; much will depend on the outcome of the SRA’s consultation. If you supervise, train, assess or educate in a firm or in chambers, I would like to know what you think those records have to hold. The window for saying so is while the post consultation position is still being formed, not after the rule is made.
On reviewing this blog and the reports it refers to, AI specialist and my mentor, barrister, Dr Peter Fields, said
“The meaningful question is not whether a lawyer has attended an AI course, but whether they can use AI competently in practice. That means understanding its limitations, verifying what needs to be verified, and being able to explain the professional judgement applied to its output. AI competence should ultimately be evidenced by what lawyers do, not simply by the training they record.”
Michaela Hardwick
