The Brief:
AI adoption among lawyers has hit 94%, up from just 41% two years ago, according to a new UK survey.
But hallucination fears are climbing too, with 83% of lawyers worried about fabricated AI outputs.
AI adoption among lawyers has hit an all-time high, but the trust problem hasn’t gone anywhere.
LexisNexis’s latest survey of 543 UK legal professionals found AI adoption has surged to 94%, up from just 41% two years ago. One in ten lawyers said they’re now dependent on AI to do their job, with that figure climbing to 15% at large law firms and 13% in-house.
Usage is increasing:
74% use AI at least once a week, with 34% using it daily. Only 6% say they don’t use AI tools at all.
Legal research is the top use case at 69%, followed by document summarisation (62%) and drafting and document review (53% each).
Surprisingly, leaders are more likely than juniors to use AI for both drafting (62% vs 51%) and research (73% vs 68%), suggesting the tech is climbing the seniority ladder rather than staying stuck at the junior end.
But the more lawyers use AI, the less they seem to trust it. 83% are concerned about AI producing fabricated or inaccurate information, up sharply from 57% in 2024. Hallucinations have topped the list of lawyer concerns since LexisNexis started tracking the issue at the start of 2024.
That’s telling. Most AI tools have introduced features that link back to source material, but clearly those features aren’t easing anyone’s nerves. It suggests the problem isn’t a lack of citations, but the quality of the underlying source material those citations point to.
The worry isn’t evenly spread either. Lawyers at the Bar are slightly more concerned about hallucinations (87%) than the wider group, and senior leaders are more worried than mid to junior staff (86% vs 82%). The people with the most exposure to risk are also the most on edge about it.
Maybe that’s not such a bad thing. A healthy dose of scepticism will look a lot smarter than blind faith when things go wrong.
The concerns aren’t just theoretical. Pinsent Masons ran into trouble with the High Court and flagged itself to the SRA after a junior lawyer relied on AI to draft letters that turned out to be misleading. In the US, Sullivan & Cromwell fessed up to AI-generated mistakes in a bankruptcy court filing this year.
Those are just a few examples of why lawyers can’t afford to take AI outputs at face value. Dr. Gerrit Beckhaus, a partner at Freshfields and Co-Head of Freshfields Lab, says the tech should support legal judgement, not replace it, with accountability for the final work product always resting on a lawyer’s shoulders.
Source: LexisNexis