Well, not sure how everyone does it, but it's hard to dispense with a basic "reading the thing and checking the sources exist and they what they are purported to say". Or having someone you trust do it. The advantage is that it should be done anyway.
I think this can be automated to some extent (and that's what I am trying to do: https://pelaikan.com/), but ultimately if one wants to master one's case, again good old reading is a must, at least until we find a more efficient way to upload information into someone's brain.
“Good old reading” for the purpose of catching AI-generated legal hallucinations can’t be done by tech people. This can only be done by highly-paid lawyers and experts in the subject matter.
This sounds like a money-losing proposition to me, because the amount of good old reading needed to catch legal AI hallucinations is enormous. I’ve done it. The economics fail to make sense.
Hi Damien! I hope you're doing well. I think you're opinion on AI-Native Law Firms is spot-on. This is something that I've been tracking as well. There is no definition of an AI-Native Law Firm. AI being embedded as part of core workflows is vague since you don't know what constitutes "core". Furthermore, there's an entire unresolved debated on the billable vs fixed-price model. The definition is something that we'll find out with time. BigLaw will be able to out compete AI-Native Law Firms because clients chose BigLaw because of reputation, brand image and trust with over years of track records closing transactions.
One of the areas in India where I see AI-Native Law Firms capturing value could be MSMEs because BigLaw has ignored businesses like export businesses, textiles, fisheries etc. These businesses have been traditionally underrepresented. However, with AI tools more value can be captured as these businesses do not have intensive legal work. Firms can make recurring revenue as these businesses tend to be stickier to law firms (loyalty) as they grow. AI-Native Law Firms could extend legal services to a set of the population that did not have access to legal services. Would love to get your thoughts on this! (It has the potential to do so)
I write a blog in substack titled "The LegalTech Thesis" where I analyze all things LegalTech. One of my posts analyzes how AI-driven due diligence tools are transforming M&A workflows. Would love to get your thoughts on this as well!
Yes, your point is exactly what I was trying to convey: they'll have to create their own demand if all this talk about being AI-first or native is ever going to pan out. Nice point that it could be a good value proposition also when targeting growing firms - not only MSMEs, but I am thinking of startups wary of going with a BigLaw (they don't see the point) / more enthusiastic for tech.
(P.S.: I subscribed to your Substack, looking forward to reading more !)
The legal system relied on signal quality, credentials, form and format. The next layer of defense to spot poor argument was adversarial and judicial review. AI "democratizes" the former and has overwhelmed the latter that is built for the pace of human cognition.
What methodology do users of legal AI use to catch hallucinations before they wreak havoc?
The reason that fake legal reasoning can’t be reliably vetted by automated processes is that it requires System 2 to spot all the anomalies.
Well, not sure how everyone does it, but it's hard to dispense with a basic "reading the thing and checking the sources exist and they what they are purported to say". Or having someone you trust do it. The advantage is that it should be done anyway.
I think this can be automated to some extent (and that's what I am trying to do: https://pelaikan.com/), but ultimately if one wants to master one's case, again good old reading is a must, at least until we find a more efficient way to upload information into someone's brain.
“Good old reading” for the purpose of catching AI-generated legal hallucinations can’t be done by tech people. This can only be done by highly-paid lawyers and experts in the subject matter.
This sounds like a money-losing proposition to me, because the amount of good old reading needed to catch legal AI hallucinations is enormous. I’ve done it. The economics fail to make sense.
Hi Damien! I hope you're doing well. I think you're opinion on AI-Native Law Firms is spot-on. This is something that I've been tracking as well. There is no definition of an AI-Native Law Firm. AI being embedded as part of core workflows is vague since you don't know what constitutes "core". Furthermore, there's an entire unresolved debated on the billable vs fixed-price model. The definition is something that we'll find out with time. BigLaw will be able to out compete AI-Native Law Firms because clients chose BigLaw because of reputation, brand image and trust with over years of track records closing transactions.
One of the areas in India where I see AI-Native Law Firms capturing value could be MSMEs because BigLaw has ignored businesses like export businesses, textiles, fisheries etc. These businesses have been traditionally underrepresented. However, with AI tools more value can be captured as these businesses do not have intensive legal work. Firms can make recurring revenue as these businesses tend to be stickier to law firms (loyalty) as they grow. AI-Native Law Firms could extend legal services to a set of the population that did not have access to legal services. Would love to get your thoughts on this! (It has the potential to do so)
I write a blog in substack titled "The LegalTech Thesis" where I analyze all things LegalTech. One of my posts analyzes how AI-driven due diligence tools are transforming M&A workflows. Would love to get your thoughts on this as well!
https://harshithviswanath.substack.com/p/ai-driven-legal-due-diligence-the
Thanks Harshith;
Yes, your point is exactly what I was trying to convey: they'll have to create their own demand if all this talk about being AI-first or native is ever going to pan out. Nice point that it could be a good value proposition also when targeting growing firms - not only MSMEs, but I am thinking of startups wary of going with a BigLaw (they don't see the point) / more enthusiastic for tech.
(P.S.: I subscribed to your Substack, looking forward to reading more !)
Damien, wondering if you are writing anything up on your read of the new US ruling on the impact of AI usage on AC privilege. https://harvardlawreview.org/blog/2026/03/united-states-v-heppner/
I wrote on Heppner there (https://artificialauthority.ai/i/187061504/the-judge-can-read-your-chat), and am collecting further decisions (there was another one last week) to create a new database tracking this issue.
Of course you did! I’d missed it.
The legal system relied on signal quality, credentials, form and format. The next layer of defense to spot poor argument was adversarial and judicial review. AI "democratizes" the former and has overwhelmed the latter that is built for the pace of human cognition.
A very good way to put it.