Despite what seems to be an accepted truism, AI hallucinations aren’t necessarily completely random. That’s the key insight from a new physics-based analysis by a group of scientists and engineers and it may change how we should be using GenAI tools.

The key finding: GenAI systems have a deterministic mechanism that causes output to flip from reliable to fabricated at a calculable step. And that step arrives exactly when a lawyer’s need is greatest. On novel, unsettled legal questions where training data is sparse.

That’s good and news. The good: if failure is somewhat predictable, more verification is needed when you are in ambiguous areas. More confidence on well-known information. 

The bad: the stretch of accurate output that precedes the failure builds false confidence by the uninformed, making the fabrication harder to catch, not easier.

My post for Above the Law.

Managing by walking around used to be standard practice. But with remote work, Zoom, and billable hour pressure, the concept lost some of its luster .

But with AI we may need it more than ever. When we rely only on LLMs to make decisions and summarize work, we lose something critical: the senior lawyer who stops by, asks questions, catches problems and offers ideas and solutions.

I had a practice group leader who walked the halls every afternoon. At the time, we all thought he was just checking to see who left early. Maybe. But in hindsight he stopped, he asked what we were up to, he listened. And he made all of us better lawyers.

Here is my post for Above the Law on why MBWA matters more than ever in the age of AI.


Deep fakes are coming to our courtrooms. They are going to change how we try cases.

Here’s what the rise of deep fakes may mean for judges, juries, and trial lawyers. Along with the impact of the so-called “liar’s dividend”: the risk that repeated exposure to AI-generated fakes causes people to disbelieve all digital evidence, real or not.

Here are three ways courts may respond. And what trial lawyers need to do to prepare. Sometimes less technology in the courtroom, not more, may be the better call.

My post for Above the Law.

asked a room full of lawyers at ABA TechShow how many had encountered deep fake evidence in litigation. Not one hand went up.

Is the deep fake threat a problem in search of a problem? Or are we in the same place we were with AI hallucinations before the first fake citation showed up in a courtroom and we just don’t know it yet?

I co-presented on this with Judge Xavier Rodriguez, one of the sharper judicial minds on technology in the federal judiciary. His observation: we have a built-in presumption of validity for photos, recordings, and video. Maybe our skepticism hasn’t caught up with what AI can now produce.

My post for Above the Law.

Two TechShow keynotes. Two very different speakers. But the same conclusion.

Jordan Furlong opened TechShow by talking about the human lawyer: the one who walks through the valley with clients, who has their back, who builds the kind of trust AI can replace.

Nilay Patel by arrived at the same place from a different direction. His argument: law is built on ambiguity and unpredictability. Software abhors both. And the best lawyers don’t just navigate the law. They navigate their clients’ anger, fear, and desires through it. Lawyers are as much therapists as they are lawyers.

Furlong and Patel: two bookends to an outstanding TechShow and a hopeful message for the future for those who are listening.

Here is my Above the Law post

Jordan Furlong’s ABA TechShow keynote was one of the best I’ve heard. His thesis: AI will commoditize legal knowledge, mechanize legal work, and reconfigure law firms. The lawyers who suceed won’t necessarily be the ones who know the most, if that was even ever the case. Instead, they will be the ones clients want in the foxhole with them when things go bad or even hard.

I’d add one thing though: the lawyers who thrive are will also be entrepreneurs. Not in the startup sense but in a more important sense: seeing a client need and doggedly pursuing a way to satisfy it.

That’s what separated the great lawyers I’ve known from the ones who just billed hours. And it’s what will separate them in the AI era too.

My full take for Above the Law on Furlong’s keynote and what I think we still need to teach young lawyers.

ABA TechShow 2026 kicks off this Wednesday in Chicago and it’s a little different from every other legal tech conference on the calendar.

Two strong keynotes (Jordan Furlong and Nilay Patel), a genuinely important Saturday session on the rule of law with three ABA presidents, 47 educational sessions, 120+ exhibitors, and the traditional startup pitch competition moderated by Bob Ambrogi. Oh, and the puppies are back. It’s a family celebration.

My post for Above the Law

When it comes to AI, all to often law firm management freeze up from too many choices and do nothing, or overcorrect and buy everything in sight. 8am’s Legal Industry Report, for example, shows nearly 75% of legal professionals are already using general-purpose AI tools like ChatGPT and Claude for work, while 71% say their firms offer zero training on responsible use and not many firms offer guidance. Firms that over correct, on the other hand, risk over exposure and non use of expensive legal specific tools they have purchased. Either way, firms need to understand what lawyers want and need before making AI decisions. And no matter what they need to provide realistic guidelines and training. My post for Above the Law.

Here’s my 2026 Legalweek recap for Above the Law: good show and good new venue (Javits instead of the Hilton). And the complaints were predictable, especially from the crowd that usually champions change. The exhibit space was a genuine improvement. The judges keynote was outstanding. The commercialization of sessions a little less so.

And btw, after summer like weather all week, on the last day, it snowed.

At Legalweek, while everyone else was talking about AI features, I sat down with Michel Sahyoun of NopalCyber to talk about something the legal profession isn’t paying nearly enough attention to: cybersecurity in the age of GenAI.

Some facts: the average time to exploit a breach is 29 minutes. AI tools can automatically and repeatedly probe for vulnerabilities. Cyber insurance that may not cover what you think it does. Backup systems that aren’t contractually required.

Law firms are particularly exposed and particularly complacent. Lawyers are bored by cybersecurity discussions. That’s a dangerous combination.

Here is my post for Above the Law.