TL;DR

Almost every lawyer surveyed by LexisNexis now uses AI in their work, and a rising majority distrusts what it produces. Worry about fabricated or inaccurate output has climbed 26 points since the start of 2024, to 83%, while the share of lawyers using the technology at all has gone from roughly one in ten in 2023 to 94% today.

Adoption is near-universal; confidence is not

The gap between those two curves is the story. Usage more than doubled in the past twelve months alone, up from 61%, and a third of respondents now reach for AI daily. Only 6% of the 543 professionals surveyed avoid it entirely.

Where it is used tracks how well errors can be caught. Legal research leads at 69%, followed by summarising documents, with drafting and review level at just over half each. In-house teams are the heaviest users — nearly nine in ten touch it weekly — while barristers sit at barely half.

Beyond hallucinations, two anxieties cluster close together: leaking confidential material and becoming too dependent, cited by 53% and 52% respectively. One in ten lawyers describe themselves as dependent already, rising to 15% inside the largest firms. Rather more striking is the 77% who think AI is leading clients and the public to misread the law, illustrated by a large-firm leader reporting a surge in claims brought by people wielding AI, not all of them with merit.

The duty has not moved

The survey lands the same week the UK Jurisdiction Taskforce published its legal statement on liability for AI harms, which answers the worry rather bluntly: using AI changes none of a professional’s existing obligations. Selecting the tool, testing it for the specific task and checking what comes out remain the practitioner’s problem. On confidentiality it goes further, treating a solicitor who feeds privileged material into an insufficiently secure system as very likely in breach.

Looking forward

Buying behaviour is already shifting in that direction. Asked what matters when purchasing legal technology, respondents ranked privacy standards, transparency and human oversight above both price and explainability. Yet under a third say AI is genuinely embedded in how their organisation operates, with 43% experimenting slowly — a mismatch between what firms are buying and what they have actually changed.