Part Three of the AI crisis series from myself and Melissa Rogozenski : The trust breakdown that’s making legal practice unsustainable. When senior partners spend evenings checking associates’ citations and local counsel can’t trust national counsel’s briefs, we’re not just dealing with verification costs, we’re watching decades-old professional relationships crumble. The AI bubble isn’t just
Like Lawyers in Pompeii: Is Legal Ignoring the Coming AI Crisis? Part Two
Every day we see another lawyer sanctioned for using AI hallucinated case citations. But the problem may not be just lazy checking. It may have something to do with economic reality.
When AI verification costs exceed savings, what happens? If it takes 8 hours to verify what AI does in 2 hours, are we actually
Like Lawyers In Pompeii: Is Legal Ignoring The Coming AI Infrastructure Crisis? (Part I)
While everyone debates AI hallucinations, we may be missing a bigger threat. The infrastructure powering AI may not be able to sustain everything vendors are promising.
Think about it: 26 major US utilities already have requests to supply an additional 711 gigawatts of new data center power. That’s nearly equal to the entire continental US
Washington Post Analysis Shows We Are Talking Too Much And Getting Questionable Advice From LLMs — And It May All Be Discoverable
A new Washington Post analysis of 47,000 ChatGPT conversations reveals a troubling pattern. People are sharing deeply personal information, getting advice that tells them what they want to hear (not necessarily what’s accurate), and creating potential discovery goldmines for future litigation.
The study found users discussing emotions, sharing PII and medical info, and asking for
Thinking Outside the Fence: What a Grade School Game Taught Me About Legal Innovation
The grade school game seemed simple enough. Grab the other team’s flag without getting tagged. But for a kid like me with not much athletic talent, the chances of being a factor other than getting quickly tagged out were pretty remote. Or so it seemed.
One of the questions I am often asked by young lawyers
The Grace To Dabble: Two Biglaw Firms Look To An AI-First Future
Two AmLaw 100 firms are doing something unusual: sacrificing billable hours to train associates in AI.
Ropes & Gray lets first-years spend up to 400 hours (20% of their requirement) on AI training. Latham & Watkins flew 400 associates to DC for a two-day AI Academy.
The revenue hit? Probably minimal. First-years aren’t profit centers
The Disco Study: A Watershed Moment Or Just More Of The Same?
New research from Disco and Ari Kaplan reveals a striking contradiction in legal’s relationship with AI and eDiscvovery. While 70% of legal professionals recognize AI’s efficiency benefits, only 35% have actually incorporated it into routine processes.
Even more telling: 42% of law firms report zero external pressure to adopt AI solutions. .
The reasons for
An Affordable AI Tool for Solo and Small Firms
Small firm lawyers keep telling me they can’t afford the AI tools big firms use. They’re not wrong, I’ve heard vendors literally laugh at affordability concerns. So when I came across Descrybe, a legal research platform with free core features (and paid plans at only $10-20/month), it got my attendtion and I dug deeper. Here’s
Mind The Gap: Why In-House Counsel Often Don’t See The Innovation They Want From Law Firms
New study from Thompson Hine: 95% of in-house counsel see little innovation from their law firms. But clients might be part of the problem. When they reward the status quo with an average 7.4% rate increases year after year while claiming innovation is “crucial,” what message are they really sending?
The gap between what clients
Get A Grip: Focusing On Risks Instead Of Benefits Is Making AI Way Too Hard
When we talk about GenAI for the legal profession we frequently focus on the risks. But Comment 8 to Model Rule 1.1 requires us to also the understand the benefits. Sometimes we make AI a little too complicated.
Two fundamental rules: Don’t put client confidences in prompts and check the output for accuracy. Here is