The duty is settled, and it runs upward
On 6 June 2025 the Divisional Court of England and Wales, Dame Victoria Sharp P and Johnson J, decided two referrals together. In one, five cited authorities did not exist. In the other, a claim for £89.4 million, eighteen of forty-five cited authorities did not exist.
The court held that freely available generative AI tools trained on a large language model are not capable of conducting reliable legal research, and that those using them have a professional duty to check accuracy against authoritative sources. Three extensions matter more than the holding itself. The duty reaches lawyers relying on others’ AI-assisted work. A lawyer may not rely on a lay client for the accuracy of citations. And leadership responsibility falls on heads of chambers and managing partners, with the court stating it will ask in future hearings whether that responsibility was fulfilled.
A fabricated citation is therefore capable of being treated as a supervision failure rather than only an individual one. That is the sentence to read to a partnership. Stated fairly, it binds England and Wales, it is a judgment rather than a study, and no reported case has yet tested a lawyer who checked competently and was misled anyway.
How large the verification burden actually is
The tempting response is that the tools have improved and this was an early problem. The measurement says otherwise. A preregistered evaluation published in the Journal of Empirical Legal Studies, running more than 200 handwritten queries against the leading commercial legal research tools, found each of the three hallucinating between 17 and 33 per cent of the time. One was accurate on 42 per cent of queries. Another was incomplete on 62 per cent.
The detail worth carrying is about length. The tool producing the longest answers, averaging 350 words against 219 and 175, had both the higher hallucination rate and the heavier verification burden. More generated text means more falsifiable propositions, and each one has to be checked by somebody who can tell.
And it is reaching courts. A curated database of decisions in which a court found or implied reliance on hallucinated material recorded 1,963 cases as at 27 August 2026, across more than thirty countries, of which 784 involved lawyers. Its compiler is explicit that this counts only instances a court addressed in writing, so it supports no rate and no trend. What it establishes is that this is documented and dated rather than anecdotal.
The firms have already started acting on the training gap
The apprenticeship argument used to be a prediction. On 27 August 2026 the Financial Times reported the largest firms acting on it, named and on the record. EY’s UK head of consulting said firms will have to reduce flexibility in order to help the human skills, and that training in empathy, storytelling and leadership was dropped during the remote-working period while AI and technical skills were prioritised. KPMG describes reinventing in-person training. Deloitte and PwC began extra coaching for their youngest UK recruits in 2023 after finding weaker teamwork and communication than earlier cohorts.
Take that for what it is, which is strong evidence of institutional belief and policy change, and not evidence of cause. Nothing in the reporting measures a skill or an outcome, the 2023 cohort effects are attributed to lockdowns rather than to AI, and office attendance is a hypothesis about the remedy rather than a tested one.
The underlying mechanism, and what actually rebuilds a rung once it is gone, is at missing rungs and should juniors use AI.
What a session looks like for this audience
Partner conferences and partnership away days, management boards and executive committees, learning and development and early-careers leadership, risk and professional standards functions, and the annual conferences of professional bodies. Forty to ninety minutes, or a closed partnership session.
The sector-specific research already exists and is published in full: how AI changes law, accounting and audit, and consulting, with the cross-professional comparison at which professions face the greatest deskilling risk.
He is the wrong choice for a legal-technology showcase, a briefing on which tools to buy, or a session whose brief is to reassure a partnership that nothing needs to change. Where the fit is wrong you will be told.
The three keynotes, and which fits this room
Keynotes run 30 to 90 minutes, in person or virtual. A 30-minute main-stage version of each talk is available for conferences; the full argument needs 40 or more. The sector examples change for the room; the argument does not. All three, with the showreel, are at keynotes, and the programme copy is at the speaker pack.
Drift versus Design
Drift versus Design: why most organisations hand their judgement to AI without deciding to, and how to design your way through instead.
Most organisations are adopting AI by drift: a thousand reasonable decisions that add up to judgement nobody chose to give away. Rahim Hirji, author of SuperSkills, shows leaders which mode they are running, where AI sharpens judgement and where it weakens it, and hands them the controls.
We Are Superheroes
We Are Superheroes: AI is the suit. The human decides.
AI makes everyone faster, stronger and more capable. It does not decide; people do. Through a family story across four generations and three continents, Rahim Hirji hands the audience the seven human skills that grow more valuable as the tools spread. They arrive thinking AI is the story and leave knowing they are.
WTH (What the Human)
WTH (What the Human): a live test of a board's own judgement.
Business is being rewired as AI arrives, and human judgement is leaving with it. In three acts, Rahim Hirji shows a board what AI is changing in how organisations run, then ends with a live test of the board's own judgement, in the room and in real time. Nobody forgets the result.
