First, 100% agreed that current models still require guidance and careful prompting. This matches my personal experience. And I agree that there will still be a big role for lawyers to play - at least for a while; I'm just not convinced that those lawyers will be employed *at BigLaw*.
Second, regarding your question: there are a few considerations to keep in mind. Lawyers are extremely non-technical, and adopting a completely new technology scares them. There is a lot of confusion in the market about which models and harnesses are the best; I couldn't live without GPT-5.x Pro, but 99% of lawyers probably haven't even heard of it. There are long-standing fears of hallucinations (which have now been almost completely solved). Some lawyers tried GPT-4 on release, realized that it sucked, and haven't realized that we're now light years beyond it. There is also a solid "nothing ever happens" component to all this: why should a boring U.S. public company adopt AI and make its outside counsel adopt AI when it wants to make a new acquisition? Finally, on the other side of the equation, there could be compute constraints (IMO, quite likely to happen although unclear in potential magnitude) that would prevent widespread use of models by the industry. It's all very messy and makes it VERY difficult to even guess at potential timelines with any kind of precision.
Agree with all your points here. I think for the legal industry generally, the two things that are absolute musts before adopting AI in any meaningful way are reliability (little-to-no hallucinations) and confidentiality/privacy.
We recently had a judgment here in the UK about a lawyer using ChatGPT to draft legal documents for court that included false information (i.e.,fake citations and authorities). Understandably, the court said that it is the responsibility of legal professionals to check documents and correct any errors before submitting them to court. But what I found particularly striking was that the judge then went on the state ChatGPT is an "open source AI tool" and so uploading documents is like placing "information on the internet in the public domain" that breaches client confidentiality obligations. I think this is a pretty wild take - even if you are using the free version of ChatGPT and have not opted out of model training, then the most you could say is that you are sharing data with OpenAI, but not with the wider public (and yes that does probably still breach confidentiality). And so obviously using the enterprise version is far more preferable for lawyers for their work (more privacy settings, inputs not used for model training, can sign data processing addendums that impose confidentiality obligations on OA), but I think this case demonstrates (a) the misunderstandings many in the legal industry have about modern AI systems, how they work and the options for configuring them (as you have pointed out) and (b) we're still clearly in the very early days in terms of identifying and crystallising the best use cases for AI, including for legal practice/compliance work, so it will be a little while before lawyers et al converge on the best ones such that they become more normalised and built with the right guardrails to proliferate widely.
Reliability is key, but I will note that, over the past few months, I have been far more concerned with reasoning accuracy than with hallucinations. GPT-5.x Pro basically almost never hallucinates in my experience - but it doesn't always correctly reason through a problem.
That's a great question. I think you're in a good place right now! After you exit law school, people won't really care about your degree that much (at least in my experience). If you do well in your first year+ in BigLaw and know how to use AI, I think you'll be positioned very well. I share your view that now is not the best time to get into a mountain of student debt if you can avoid it - both because you already have a BigLaw job and because having less debt will mean that you'll be much more flexible. Being as flexible as possible during times of great change is IMO very important!
I'm far from convinced that private practice jobs will disappear for junior attorneys - and especially those who know how to use AI. BigLaw might disappear at some point (which is the point of my article), but I think it will be replaced by more in-house jobs and jobs at smaller boutique law firms. (I could be wrong about this, of course; AI is very different from any other technology!) On balance, I'd probably stay away from working in the government unless you WANT to work in the government. That path might seem safer, but on the other hand the government could also do random layoffs as its legal functions are automated by AI, and those layoffs would be less likely to be based on merit than a layoff in a law firm (where being able to use AI could be one of the things that keeps you safe even in the worst-case scenario, assuming that you also otherwise do good work).
I would comment on one thing you say though: "GPT-5.x Pro knows the tax laws and the regulatory implications. I also think we’re not far away from an AI harness that would enable a SOTA AI model to succinctly (and fairly quickly) summarize all legal implications of a particular fact pattern (e.g., in a legal memo) or implement them via contract language." I think current models still require quite a bit of guidance and careful prompting when it comes to interpreting legal provisions correctly and applying that to fact patterns. I saw this video (https://www.youtube.com/watch?v=T4R4qhG-RyE) from a barrister giving different models typical legal scenarios to work through with varying results. So I certainly think there will still be a big role for lawyers to play in providing the right scaffolding/harnessing/context engineering to make sure the models at least start off in the right direction when solving legal problems or carrying out a certain task, and this is in addition to the verification layer that will need to be there too.
But otherwise, I would broadly agree with the implications you highlight here. One question I would for you is what would be the human frictions that would stand in the way of this AI disrupting BigLaw in the way that you describe? The legal industry is very risk-averse/conservative and highly resistant to technological advances that require lawyers to change the way they practice? Do think this will have a significant impact on the pace of the impact from AI, perhaps adding several years to it? Or do you think the gains from AI are too great and the impact will happen quite quickly?
Hey Prinz, great piece that aligns with many of the concerns I have. As an aside, I'm a student at a top 20 law school and its been astounding how many classmates just refuse to use AI, even when its allowed. I'd venture to say that a majority of the class is AI skeptical, either because they doubt its capabilities (usually they've only used free, non thinking models) or for misfounded "ethical concerns" (e.g., the vastly overstated and mischaracterized water usage stat), although I'm at a very left learning institution. It really worries me how unprepared many of these classmates are for the future.
On that note, I'm at the top of my class. I have a great big law offer I accepted for this summer. I'm also considering transferring to a top 3 school. I have a full scholarship at my school I'd pass up to transfer. My question for you is, under the circumstances, what would you do? I have to think a top 3 degree would be more "AI proof," and would open doors in other industries, but I'm also worried that if the number of jobs (or their pay) precipitously declines, I'll be left deep in student debt. Additionally, in light of these concerns, do you figure a government job (probably at the state level) would be a better career path for a new attorney? I figure that sort of job is easier to get now, than in 5-10 years when all the laid off big law attorneys come knocking. And generally, would love to hear your thoughts and advice for current students. (I would advise law students applying to law school next cycle to simply not apply, at risk of sounding a doomer)
First, 100% agreed that current models still require guidance and careful prompting. This matches my personal experience. And I agree that there will still be a big role for lawyers to play - at least for a while; I'm just not convinced that those lawyers will be employed *at BigLaw*.
Second, regarding your question: there are a few considerations to keep in mind. Lawyers are extremely non-technical, and adopting a completely new technology scares them. There is a lot of confusion in the market about which models and harnesses are the best; I couldn't live without GPT-5.x Pro, but 99% of lawyers probably haven't even heard of it. There are long-standing fears of hallucinations (which have now been almost completely solved). Some lawyers tried GPT-4 on release, realized that it sucked, and haven't realized that we're now light years beyond it. There is also a solid "nothing ever happens" component to all this: why should a boring U.S. public company adopt AI and make its outside counsel adopt AI when it wants to make a new acquisition? Finally, on the other side of the equation, there could be compute constraints (IMO, quite likely to happen although unclear in potential magnitude) that would prevent widespread use of models by the industry. It's all very messy and makes it VERY difficult to even guess at potential timelines with any kind of precision.
Agree with all your points here. I think for the legal industry generally, the two things that are absolute musts before adopting AI in any meaningful way are reliability (little-to-no hallucinations) and confidentiality/privacy.
We recently had a judgment here in the UK about a lawyer using ChatGPT to draft legal documents for court that included false information (i.e.,fake citations and authorities). Understandably, the court said that it is the responsibility of legal professionals to check documents and correct any errors before submitting them to court. But what I found particularly striking was that the judge then went on the state ChatGPT is an "open source AI tool" and so uploading documents is like placing "information on the internet in the public domain" that breaches client confidentiality obligations. I think this is a pretty wild take - even if you are using the free version of ChatGPT and have not opted out of model training, then the most you could say is that you are sharing data with OpenAI, but not with the wider public (and yes that does probably still breach confidentiality). And so obviously using the enterprise version is far more preferable for lawyers for their work (more privacy settings, inputs not used for model training, can sign data processing addendums that impose confidentiality obligations on OA), but I think this case demonstrates (a) the misunderstandings many in the legal industry have about modern AI systems, how they work and the options for configuring them (as you have pointed out) and (b) we're still clearly in the very early days in terms of identifying and crystallising the best use cases for AI, including for legal practice/compliance work, so it will be a little while before lawyers et al converge on the best ones such that they become more normalised and built with the right guardrails to proliferate widely.
Agreed with you here as well.
Reliability is key, but I will note that, over the past few months, I have been far more concerned with reasoning accuracy than with hallucinations. GPT-5.x Pro basically almost never hallucinates in my experience - but it doesn't always correctly reason through a problem.
That's a great question. I think you're in a good place right now! After you exit law school, people won't really care about your degree that much (at least in my experience). If you do well in your first year+ in BigLaw and know how to use AI, I think you'll be positioned very well. I share your view that now is not the best time to get into a mountain of student debt if you can avoid it - both because you already have a BigLaw job and because having less debt will mean that you'll be much more flexible. Being as flexible as possible during times of great change is IMO very important!
I'm far from convinced that private practice jobs will disappear for junior attorneys - and especially those who know how to use AI. BigLaw might disappear at some point (which is the point of my article), but I think it will be replaced by more in-house jobs and jobs at smaller boutique law firms. (I could be wrong about this, of course; AI is very different from any other technology!) On balance, I'd probably stay away from working in the government unless you WANT to work in the government. That path might seem safer, but on the other hand the government could also do random layoffs as its legal functions are automated by AI, and those layoffs would be less likely to be based on merit than a layoff in a law firm (where being able to use AI could be one of the things that keeps you safe even in the worst-case scenario, assuming that you also otherwise do good work).
Really interesting perspective.
I would comment on one thing you say though: "GPT-5.x Pro knows the tax laws and the regulatory implications. I also think we’re not far away from an AI harness that would enable a SOTA AI model to succinctly (and fairly quickly) summarize all legal implications of a particular fact pattern (e.g., in a legal memo) or implement them via contract language." I think current models still require quite a bit of guidance and careful prompting when it comes to interpreting legal provisions correctly and applying that to fact patterns. I saw this video (https://www.youtube.com/watch?v=T4R4qhG-RyE) from a barrister giving different models typical legal scenarios to work through with varying results. So I certainly think there will still be a big role for lawyers to play in providing the right scaffolding/harnessing/context engineering to make sure the models at least start off in the right direction when solving legal problems or carrying out a certain task, and this is in addition to the verification layer that will need to be there too.
But otherwise, I would broadly agree with the implications you highlight here. One question I would for you is what would be the human frictions that would stand in the way of this AI disrupting BigLaw in the way that you describe? The legal industry is very risk-averse/conservative and highly resistant to technological advances that require lawyers to change the way they practice? Do think this will have a significant impact on the pace of the impact from AI, perhaps adding several years to it? Or do you think the gains from AI are too great and the impact will happen quite quickly?
Great post overall!
Hey Prinz, great piece that aligns with many of the concerns I have. As an aside, I'm a student at a top 20 law school and its been astounding how many classmates just refuse to use AI, even when its allowed. I'd venture to say that a majority of the class is AI skeptical, either because they doubt its capabilities (usually they've only used free, non thinking models) or for misfounded "ethical concerns" (e.g., the vastly overstated and mischaracterized water usage stat), although I'm at a very left learning institution. It really worries me how unprepared many of these classmates are for the future.
On that note, I'm at the top of my class. I have a great big law offer I accepted for this summer. I'm also considering transferring to a top 3 school. I have a full scholarship at my school I'd pass up to transfer. My question for you is, under the circumstances, what would you do? I have to think a top 3 degree would be more "AI proof," and would open doors in other industries, but I'm also worried that if the number of jobs (or their pay) precipitously declines, I'll be left deep in student debt. Additionally, in light of these concerns, do you figure a government job (probably at the state level) would be a better career path for a new attorney? I figure that sort of job is easier to get now, than in 5-10 years when all the laid off big law attorneys come knocking. And generally, would love to hear your thoughts and advice for current students. (I would advise law students applying to law school next cycle to simply not apply, at risk of sounding a doomer)