Judge Rakoff does NOT want you using AI, lawyer anon
This week: Depending on which corners of the internet you lurk in, you may have heard that the inevitable march of AI will destroy life as we know it, routing the stock market and leaving you a permanent member of the underclass destined for a life of delivering food across a panoply of vibe-coded apps. I’m not here to say whether this is an optimistic or pessismistic take, but what I do know is that anyone who spills pixels trying to set out what the world will look like as a result of AI adoption is skipping a beat or two. When we can’t even predict what AI adoption will look like from year to year, it’s a bit difficult to say anything meaningful about its second-order effects. In the legal sector, beyone the uncertainty around technological developments, you have the impossible-to-predict outcome of all the pending discovery motions over privilege and AI that will eventually work their way through Appellate Courts, and which will ultimately shape the direction of what AI tools are used and how they are implemented. Can anyone say “on-prem local models”?
Against this backdrop, let’s look at what legal adoption of AI looked like in 2024. Per the ABAs 2024 Artificial Intelligence TechReport:
Of the leading AI-based research tools that firms have already adopted or are seriously considering adopting, the top three platforms cited overall were ChatGPT (52.1%), Thomson Reuters CoCounsel (26.0%), and Lexis+ AI (24.3%). ChatGPT was the clear leader across firms of every size, while the remaining market share fluctuated with respect to the other two most widely used tools and varied depending on firm size. Digging deeper, the following platforms were name-checked only by respondents at firms with fewer than 50 attorneys: Westlaw AI, Co-Pilot, Summize, DocDraft, and Alexi. Conversely, only individuals at firms with 50 or more lawyers cited Paxton AI, Henchman, Blue J Legal, and Robin AI as part of their current or prospective tech stack.
From the same report, we see the following usage statistics: 30.2% of attorneys indicated that their offices were currently using AI-based technology tools, with reported usage rates running the highest within firms employing 500 or more lawyers at 47.8%. Adoption rates drop off sharply at 29.5% for firms with 10-49 lawyers and continue to fall for firms with 2-9 attorneys (24.1%) and solo practitioners (17.7%).
Looking at the trend through 2025, the ACC/Everlaw GenAI Survey found that corporate legal AI adoption more than doubled in a single year—from 23% in 2024 to 52% in 2025. It's not surprising to see corporate legal use of AI tools outpacing firm usage. For now there is probably less fear of privilege concerns, until the first opinion comes out that results in an entire privilege log getting turned over to the other side in discovery because the legal department turned everything over to Claude. Until then I will keep letting Claude spelunk and boondoggle for me as much as I can.
It was a busier week than I expected so here is some AI slop for you, dear reader, of some AI accumulated statistics and presentation reviewing the trends from various reports of perhaps questionable authenticity.
The slop does allow you to tell some interesting stories that seem to align with real-world common sense at least. One of the more interesting takeways was that use of general purpose legal tools has increased while legal-specific tool use has decreased. This matches my use patterns and is a trend that will probably only continue (and I suspect the amount of “off the books” use of Claude/ChatGPT/Gemini on their own accounts, if accounted for, would make these numbers even higher).
Other news:
Ryan Peterson (Flexport) threw some shade at Harvey for some reason. He is probably right though and Claude will eat the opportunity for a lot of legal AI startups.
The reports of lawyers being sanctioned for AI use are coming non-stop. I’m going to eventually devote a post to digging on the details of some of these.
