I wonder if it's really that rare for lawyers to look at a case from a strategic / economic point of view? When at some point in the past I was considering suing a company for breach of contract, that was pretty much what the entire initial conversation with our lawyers was about: they considered the timeline, probability of winning, costs for the lawyer'ing, costs for external experts, ability to recoup legal costs if successful or to incur those of the counterparty if not, ability of the counterparty to pay, alternatives to litigation and so on.
After about a day of prep work on my part, having them read through that prep work for about an hour and then discussing it with me for another hour, we concluded that the expected outcome was break even, with a time investment that was decidedly not break even, and did not proceed to litigation. They really helped our company to look at the problem objectively and without emotion, and I really hope that's not a rare experience!
(That said, I do understand that strategies for successfully dealing with being sued might need to get a bit more creative than those for deciding whether to sue.)
I think this article really buries the leadâwhat really allowed him to win was leverage. He says thatâs an over beers info, not a blog post. But thatâs actually what let him winânot learning from the LLM how to understand his lawyers arguments better.
Developers and lawyers have a lot in common. In the end, we write rules that are interpreted, whether a machine or a judge.
Therefore I think the same things apply as in software development, when you start to think you can fire lawyers and replace them with AI: you will then bear the responsibility of the job they did before. That means it can go the right way, but it can also go the wrong, wrong way.
In the US, our system of lawyers is broken. If we are to be a People governed by laws and not persons, then we People need to be able to use the law to right wrongs. In other words being able to use the law by using the courts is a requirement in a democracy.
As an experiment try to go to court pro se. The experience is characterized by the fact that in US courts, pro se clients can be sanctioned for improper actions, but lawyers can not be sanctioned for purposeful improprieties by pro se clients.
It is sort of like a commoner going up against a knight. The knight has armor and a lance and you have your scythe. Fun! In other words it is rigged.
And here is the kicker. Lawyers are "officers of the court". You could take that to mean that they uphold the standards of the court. Or when a lawyer does bad things, that could mean that the other lawyers - including the judge - see that lawyer as one of them. You know, sort of the way that many priests see abusive priests as priests first and abusers second, and so take actions to "protect" the church.
So whats a poor person to do? You try to get a lawyer. Lawyers will not want to take you on if the other side has a high power lawyer and lots of money. Not because your case isn't proper, but because the economics are not there. For the lawyer. At some point the side with more money will simply run you out. Your lawyer will urge you to settle.
Again this is not an us vs them thing. The issue is that if common people cannot afford to obtain justice, then we are not a democracy. We are something else. Corporations and the ultra wealthy have deep pockets to use the courts. If we tolerate that, it is reasonable to expect that the common people will not find this kind of "democracy" to their taste.
Thank God I've never been involved in serious litigation, but his description of the process sounds absolutely wild to me. Reminder that the US is a global outlier in 'loser doesn't pay winners' attorneys' fees' (as he notes, this is literally called the American Rule). And a global outlier in using juries for civil trials, which can lead to some gigantic unpredictable judgements. I'd imagine our discovery laws are unique too.
Surprised some business-friendly US state hasn't greatly restricted discovery in civil litigation, so as to attract companies to incorporate there. And this is why arbitration rose as a way to settle disputes- if you look at the standard arbitration agreement they greatly limit discovery
This article has good advice but itâs not really AI advice thatâs the important tip.
The main observation is you 100% have to manage lawyers aggressively and understand and question what they are doing at literally every step or the budget will be out of control. Sounds like AI helped this guy but itâs more about your personality/approach than anything else.
Also donât hire $1000 an hour lawyers for litigation unless youâre a major corporation in a complex case with a ton at stake.
> Upload Everything to AI Projects.
Unless you're using an ephemeral on-prem LLM that doesn't keep logs you should begin with an assumption that your AI chat transcripts will be available to your opponents in discovery. There isn't clear caselaw on this yet, but given that material you share with third parties other than your lawyers almost always is and sharing your strategy musing would probably be disastrous, you should plan accordingly.
Particularly, subpoena are quite powerful in the US and can be obtained quite easily and often without an adequate opportunity for the ultimate user to oppose them. The AI providers will comply with them as just a matter of process, and the first notice that an opponent is seeking your transcripts may be after they already have them. (And that's assuming your opposition doesn't have insiders at the AI company to begin with...)
That aside-- Having thoroughly dispatched an AI dependent opponent in court, uhh, well lets just say that I hope any future opponents read this article. ChatGPT outputs an astonishingly embarrassing amount of completely wrong pseudolegal nonsense and set him up for multiple case losing disasters.
There are also a number of advanced strategies available to an opponent who has realized you are using an LLM, as they now have oracle access to something similar to your own AI advice and can test their actions against the advice it will cause you to get from your AI. ... even so far as brute force searching rewrites of the language in their correspondence to increase the odds that you get disastrous advice.
If there is no attorney-client privilege with AI (I believe there is not), then could the opposing party demand discovery of all of oneâs communications with the AI pertaining to the case?
Watched an expat video on youtube recently where he said he would never go back to America because the legal system is effectively lawless. Reading things like this, and others, makes me agree. There is this idea in America that countries with Civil Law systems instead of Common Law are more dangerous, but are they really more dangerous than "the law doesn't actually mean what it says, it means what this judge said 80 years ago it says"? The financial asymmetry just makes it worse, wealthy people and entities can very easily lawfare their enemies out of existence.
I used ChatGPT Project folders extensively to deal with my landlord and the rental contract due to problems with the windows. Things went from the landlord saying "maybe they'll be fixed in 2 years if you extend your contract" to "I've called a contractor and am waiting on a quote, should be done in a couple months". I'm not well versed in UK law or how to speak/write "properly" nor have the fortitude to parse long contracts and laws, ChatGPT handled it well enough for me to get what I wanted.
There's a lot of mental cycles being expended on "how to avoid going to court over contract litigation" when the answer could be provided by a bog-standard forum selection clause requiring arbitration from a template last updated in 1992 or indeed a chatbot; but you would have to ask it the right question.
Looking up the lawsuit in question but without reviewing the record, it looks like this was probably as much personal as it was business. This always makes cases a lot harder to settle. When it is just business, you can almost always work it out in numbers. When you screw over a wealthy man in a very personal and humiliating way, your contract case can become like divorce, which is only good for the lawyers and for no one else. Ask the chatbot to summarize "The Prince" for you and maybe this will be one of the points that it gives you.
I think the major point is that if you fancy yourself a beginner businessperson, sooner or later you will have to find some professionals you can trust. Regardless whether that's a lawyer or an accountant. Doing business will have its costs, and you will not necessarily pay those costs in the way you can anticipate.
Maybe you pay those costs before you sign a contract, and avoid a jurisdiction in a contract that you have no connection to (if you can). But you will definitely pay these costs when you have to litigate. If you think Claude or ChatGPT is the best answer to this, and all you need to know is using advanced voice and upload documents, I believe that's not that useful piece of advice.
Please bear in mind that it's not just particular case law that these tools hallucinate. They give incorrect advice in very different ways how humans do. If you can find a lawyer you can trust in a jurisdiction you're interested in, and you got along with him/her quite well, you can trust them for as long as they're in business, even if you never pay them once again. And they might even give you referrals or point to other professionals, understanding your budget and situation. Of course, they can still give you bad advice, they can become gambling-addicted, and run away with your money. It happens, even if very rarely.
But with LLMs, you won't know when they hallucinate, and what they get wrong. Even if you have the full statutes and case law uploaded and updated, it will give you wrong answers as well, but they're not working in the same way as humans do.
All we know that it's more likely to give you incorrect advice in jurisdictions not being covered by their training data (like outside New York or Delaware), and it will draw incorrect inferences in other cases from materials it has seen. Usually only lazy lawyers not reading the output will face the problem of incorrect case law.
But legal advice is not like coding where you can filter out most of the incorrect answers with tests and compilers. The author have happened to chose the wrong lawyers for the first time. I accept that not everybody has reliable lawyer friends qualified in Delaware, but in terms of advice, it would have been more practical to discuss the possible directories to use, beginner's methods to select lawyers, warning signs to look out for etc. It's sad that many lawyers act for trolls and do bad faith litigation. It's even worse that bars have no methods to exclude and disbar such guys. But you cannot avoid these problems by trusting a cheap AI tool instead, that's just wishful thinking.
An interesting post by Tyler Tringas about being sued as a founder. Obviously there's an AI angle to this post, but I also found it an interesting look in to the power dynamics of civil litigation in the US.
In a previous post [0] they talk about the impact of the litigation, which seems to have contributed to the shuttering of his fund
Simply put, this person hired the wrong law firm.
When you get sued, you get a reasonable amount of time to respond. That includes time to interview firms, discuss the approach and goals, and then choose the best firm that will attempt to carry it out. And then that firm will have enough time to write and file the response. Of course, you have to start immediately.
A good litigator wants to give the other side their bar number, and have them shit their pants when they look up their litigation record. Itâs absolutely a strategic advantage; milking hours and aiming for a settlement is bad for a litigators career.
Did I somehow miss the conclusion of the piece? I was hoping to finish with reading about how AI found that âone simple trickâ to get the case dismissed or won and also recover costs.
A friend of mine is regularly mentioning chatGPT as is lawyer.
Given many countries share the same language, I am struggling to understand how he make sure chatGPT is not hallucinating and that he is basing his advice on the laws of the correct country.
> The Delaware legal system is fundamentally broken for defendants. Due to the ironically labeled âAmerican Rule,â even if you win, itâs extremely rare for the other side to have to cover your attorneysâ fees
Another reason not to blindly use Delaware
There are 49 other states, a district, territories, and various native reservations, with their own incorporation statutes
They havenât all just sat idly over the last 30 years in the competitiveness of their statutes
Some even have Chancery Courts just like Delaware
And the lack of case law is actually a benefit. If you donât like Delawareâs case law you can argue for a completely different ruling that only benefits you. while if you do like Delawareâs case the local judge has a strong incentive to lean on it and you can push for that.
Its surprising how uninspired people are about this, especially ones that arent raising capital
> Some lawyers bristle at clients setting strategic direction. They believe thereâs a âcorrect wayâ to do things, they know it, and theyâll fight you if you disagree. In that case, get different lawyers. Ultimately I was able to find good counsel who were willing to work with me taking an active role.
Brrrrrr. AI an be enormously helpful when it comes to trying to understand - I have no doubt the author got exactly what they needed here. But if you're making blanket rules about ditching expert opinion because you can wield a model then I think you're letting your own engineer's syndrome drive things, and you're letting it be validated by survivorship bias.
I don't know that this is a particularly good article, but it touches on two elements of games based on incomplete / imperfect information.
One of these elements he refers to as "leverage" and I think he does an ok job explaining what it is, although maybe not how / why it works.
The other one, contrasting doctors and contractors, I think he does poorly. Full disclosure: I've had really crappy luck with doctors (it's not just luck). Doctors aren't used to getting "not to exceed" demands, things like that. On the other hand my father was a psychiatrist, a specialist, and that gives me insight into the "practice" of medicine; and I've worked for a number of biotechs and I have some insight into the evolution of medical practice. Combined, I have a pretty good sense of when the doctors are "practicing" versus forming a hypothesis (ask for one!).
During one of my burnouts I took a "vacation" working (what started as part time) for a high-end flooring contractor. Although the owner and I had different backgrounds and beliefs we had mutual respect for each other (and we had great conversations). But what the public calls (and in public we called) an "estimate" we called internally an "interview"; and we wanted to take the temperature of the "pain in the ass" factor: Does the prospect truly understand the work? Does the prospect have irrelevant concerns? (I'm doing a poor job here, although I'm pretty sure successful contractors will understand what I'm talking about.) Our bids could vary as much as 300% (upwards) from our baseline model / estimate for the job... before we said "no" outright or came up with good excuse for them to fire us before receiving our bid.
There were some rather "zen" aspects of the process (my father had a zen question: how do you feel? I have a zen question: what do you measure?), such as an insistence that all special instructions had to fit, legibly handwritten, in about 1/3 of an 8x11 piece of paper. (We did $10000 floors on a one page contract, not including non-negotiable required boilerplate; why can't we do that in IT? Can we? Yes, yes I say we can.)
But distilling this down: some contractors would rather deal with the masssages and fish tank cleaning (at 3x the $$$) than do floors... whereas for us flooring was a religious calling or something. You'll find this both in doctors and in contractors; and you'll also find people out there who expect to pay that 300% markup to have the contractor walk their dog.
Amazing blog post.
I'm new to entrepreneurship and just signed my first few corporate legal agreements. I was also deeply aware that anything agreed to was basically a gentleman's agreement--it wouldn't be worth either party going to litigation over, and even less so if the other party was internationally based.
So I've been following the "no-assholes" policy as well.
Overall, this seems like a good take. However, I've never really seen this:
"When you tell lawyers âwhat do you think we should do,â theyâre going to give you the âbest representationâ they canâbecause thatâs what theyâre incentivized to do. This means ticking every box, being as thorough as humanly possible, and taking everything the other side throws at them as seriously as possible."
My experience with smaller matters seems to be that they don't give it much attention because it's not worth their time. I think the author's take only applies above the small claims level.
I don't think these lawyers were trying to give their "best representation". They have seen enough cases to know exactly what the best representation would look like and that this wasn't it. But lawyers are like business consultants. Poor work costs them nothing. In fact, it earns them more since they are retained for longer. This is a world where the lawyer needs to actually have an incentive to solve your problem as quickly as possible, and that is extremely rare. Having skin in the game could be one way.
> Donât take the AIâs word for itâconfirm everything.
This is all you have to do to counter kneejerk AI naysayers
You can also do this when talking to professionals its just that those professionals are charging you hourly, have low emotional intelligence and dismissing your ability to comprehend anything, have an ego to defend the discipline they had to use to get there, and confirming anything with them requires a separate meeting weeks later
Case law hallucinations and even law hallucinations are very dangerous. Any made-up citation can easily unravel your entire argument, and is a great example of how AI safety needs to be experienced to be taken seriously. The advice to use AI as a guide rather than a lawyer is crucial to follow when a million dollars is on the table.
> What lawyers typically wonât do is come up with creative ways for you to accumulate leverage against the other side to get them to back down.
Unfortunately, it seems you chose the wrong litigator. This, in a nutshell, is the job.
I would not invite the legal risks involved in giving a third party access to every business contract I have. Writing contracts in a way that allows for this and mitigates all or most risks is likely quite expensive.
I've always viewed lawyers as (first) professional negotiators and (second) experts on law. I'm surprised that the author had to take the lead on the negotiation part.
> This forced me to develop practices of radical acceptance and get comfortable with circumstances I couldnât change.
When you are in litigation you are at war, embrace the suck.
Thank you, that was an interesting read and gave me a perspective on something I knew nothing about
Use "AI" to fix problems caused by a broken legal system.
Why would you fix the legal system instead...
This is excellent! I hope big companies read this and think twice before trying to: a) take advantage of less-litigious players; b) create abusive contracts boobytrapped to go to court or force a settlement for control/IP/free-service later.
I hope it spurs a larger "moral reckoning" movement of legal strategy among "big players" - the moral depravity or trying to abuse other innovators through deceptive legal gaming is an illness US corps should cure themselves of.
My advice to developers/engineers who start their own corps is: always read every contract, think through the implications, consider worst cases, and only sign ones you're comfortable with. If you don't like anything, push back and negotiate. How does the other side they come off in that process? Make note of who they show you they are and incorporate that knowledge into your ultimate decisions.
I guess if you're VC backed the calculation is different: let the advisors and LPs absorb the risk and handle it for you. But if not: you need to get involved deeply. AI is a superpower that hopefully stops the abuses that have been so rampant. You got this!
> 1: Upload Everything to AI Projects. Use the projects feature in Claude or ChatGPT to upload all relevant documentsâcontracts, motions, emails, anything relevant to your case. You might need to be selective if you hit context window limitations with thousands of pages, but include as much source material as possible.
> For what itâs worth I didnât necessarily find one platform (Claude or ChatGPT) or even specific models to be materially better than others at this work. ..
> 3: Leverage AI for Contract Analysis. AI excels at reviewing lengthy contracts. When you have a 400-page agreement where page 6 references âcustomary exceptionsâ that are finally listed on page 421, AI can instantly find and connect these references without you having to jump around through table of contents.
> Always have the AI excerpt the actual contract language with page numbers, then verify by searching the document yourself. Donât take the AIâs word for itâconfirm everything.
Check out Tritium (https://tritium.legal) to save this upload step. Open the folder (or right-click your folder on Windows) with your documents and chat with your model of choice.
I am not an attorney, but have been a litigation legal admin for over two decades in a major American megacity. I feel that this position may actually lend me towards a more objective analysis than a litigation attorney or the article's author, because I am a longtime assistive witness to the process of law without necessarily being its direct proponent/creator.
There were things I found problematic about this article. One of the primary aspects I found incomplete was that it doesn't go on to specify HOW he supposedly used AI to avoid the necessity for a litigation attorney. To me that was a vital piece of information necessary for the author to prove their point, and its omission means the essay is significantly flawed.
Much like a surgeon, a litigation attorney knows 'surgical' techniques - they know the specific dynamics of law that are at play. They further know from local experience with judges (and their colleagues' experience with same) and the "plaintiffs' bar" (i.e. people who are looking to make cases) which techniques are best for which set of circumstances, much like a surgeon can evaluate which surgical techniques are best suited for the job they see in front of them -- they have the power of accurately observing and evaluating what is in front of them in full 3D high definition, whereas AI is going to be subject to how the facts are reported to it (and what facts may be either purposefully or accidentally omitted by the reporter).
None of those techniques are gone into by the author; he merely indicates that he used AI, without going into what it advised him to do. The act of omitting that piece of information means - at least IMO - that the editorial itself cannot by definition make its point.
Much like a surgeon, a litigation attorney knows 'surgical' techniques - they know the specific dynamics of law that are at play. They further know from local experience with judges (and their colleagues' experience with same) and the "plaintiffs' bar" (i.e. people who are looking to make cases) which techniques are best for which set of circumstances, much like a surgeon can evaluate which surgical techniques are best suited for the job they see in front of them -- they have the power of accurately observing and evaluating what is in front of them in full 3D high definition, whereas AI is going to be subject to how the facts are reported to it (and what facts may be either purposefully or accidentally omitted by the reporter).
None of those techniques are gone into by the author; he merely indicates that he used AI, without going into what it advised him to do.
I will agree that - much like it has done for me with medical knowledge - AI has the power to "prep" you very well for appointments with your attorney, assuming that it is not giving you a higher-level hallucination or leading you into a knowledge blind spot, which, as amateurs in the profession, laymen may not know enough to recognize (I don't omit myself - I would at most consider myself a 'talented amateur' at law).
But entirely omitted from all this evaluation: attorneys nowadays are very conscious of a client's desire to be budget-conscious; they know they charge an arm and a leg and they will try not to, and will try to help. They're not always just out for the almighty dollar, despite the caricature. They will often know what you are looking to spend and they will work to accommodate such price range - because they want to foster client loyalty to them or to their firm, and garner a favorable impression and word-of-mout. Just as one example, as a very experienced legal admin, I have sometimes done things free (not illegally so) that might have, in old days, been done for cost by an associate. (I am not speaking of the practice of law. But attorneys in olden days might've had associates charge for administrative things that can be done freely, and with greater ease now with the presence of the Internet.)
Also, one reason why lawyers are more like surgeons than they are "general contractors" is that lawyers specialize. There are multiple fields of law expertise, much as there are multiple fields of surgery, and litigation attorneys are a particular surgical field.
In short, while I agree with some aspects of the essay - I think that there's value in allowing AI briefing to get you up to speed to be a more educated client, which itself can then save you time with an attorney, which may then translate to money saved - I think there are multiple statements within the article that are simply flawed, wrong, or don't reflect the current reality of law, and how lawyers interact with clients nowadays.
None of the above reflects any attorney-client confidential information. I am not a lawyer, I am not your lawyer, and I do not speak for (nor am I empowered to speak for) my employer.
As engineers we are (rightly) very scared of AI taking our job.
If I were a lawyer I would be even more scared.
> The Delaware legal system is fundamentally broken for defendants.
The entire US legal system is fundamentally and utterly broken.
If you want to make a few thousand dollars, I hate to share this secret thing you can do, but the reality of my experience is that the only thing stopping more people from doing this is, you know, some semblance of morality. Go find any local, smaller nonprofit organization. Volunteer with them for a day or two. Pay $80-$150 to sue them in small claims court for $5000 because <<you can literally make up any reason, you don't need evidence, it doesn't need to have happened, "mental hardship", I'm actually wondering if you hadn't have needed any interaction at all with the organization>>. Have an LLM generate 2-3 emails a week outlining anything at all related to grievance; each one you send to the organization is likely to result in billable legal hours. When the court date gets near, you can file for a motion to extend. Keep sending emails. After 4-6 months of doing this, send them a settlement offer for half of what you originally sued for.
I've seen this essentially destroy one nonprofit; the board was inexperienced in dealing with it, the person doing it had clearly done it before, by the end they'd spent ~80% of their bank account on legal fees, paid out a few thousand in a settlement to the plaintiff, and half the board resigned in burnout.
If you've never came into contact with this kind of weapon before, you might have it in your head "we'll just argue the case, we're clearly in the right" ha. ha. No. You'll never get the opportunity, and the people who do this know it.
If you're now thinking "well, if you're such a small organization why are you paying lawyers at all? We have AI. We'll just file responses on our own" I wish it were the case. But sadly: at least in my state, you can sue organizations in small claims court without representation, but filings made to the court in defense of the organization must be made by a licensed attorney. Asymmetry is built into the system.
The only thing that can protect nonprofit organizations from this, beyond having a massive bank account, is getting a lawyer to donate work pro-bono and/or sit on your board of directors. This is a common thing for nonprofits, for this exact reason, and assuming you're legit and well-known its not too hard to find legal experts who will do this for you.
If you're not a nonprofit, I have to imagine you don't have this angle open to you, and you're just SOL.
âI used imperfect LLMs to âwinâ (not lose) an imperfect lawsuitâ
I highly encourage people to follow this advice. My fellow lawyers could use more people like him relying on AI instead of a lawyer. It makes it sooooo much easier for opposing counsel to win cases that would otherwise have been a slogfest against competent counsel.
Really, it fundamentally boils down to this: the CEO didn't bother tell his legal counsel he wanted to minimize legal spend. He just told them to do their thing. So they did, and because they weren't given constraints they did their thing well. But it wasn't what the CEO actually wanted them to do. The problem was that he didn't tell them what he wanted to do, and assumed that lawyers were telepaths and knew what he wanted.
So he did a dumber thing and decided to turn to an LLM, instead of doing the thing he should have done in the first place: tell his original lawyers what he actually wanted. He ended up getting lucky and not getting burned by it due to facts he chose not to disclose, but he is the rare exception.
Also, there are some huge falsehoods in his post, like #3. AIs are absolute shite at reviewing lengthy contracts. They're great at summarizing things where the correctness of the summary doesn't matter, but they're horrible at summarizing things where being correct is important. Also, his example of how the AI is used isn't AI. It's literally just a keyword search, which has been a thing for 3 or 4 decades now, and has been a basic function of doc review software for the last 2 decades.
#4. No, just no. It would cost more money to have the lawyer review an AI-generated legal document than to create one from scratch. Lawyers don't actually create most legal documents from scratch; they use templates from contracts and pleadings that have already gone through the legal trial by fire. So you're actually increasing your legal spend quite dramatically if you use AI instead.
Hallucinations. New models are not getting better at eliminating hallucinations; they're actually getting worse. Many of the recent AI-related court fails were from lawyers using the LLM tools provided by legal service providers like LexisNexis, and those LLMs were trained solely on legal databases. Even a single hallucinated citation in a legal filing allows the judge to void the entire filing and sanction the lawyers and client; in many situations the court this can result in a procedural loss for the client that they would have otherwise won on the merits if it had proceeded to trial.
TLDR: use AI if you want to make things easier for the other side of the lawsuit. Don't be like Trump. Experts exist for a reason. You wouldn't ask your lawyer to program your backend, so why would you turn to an AI for legal advice?
Excellent blog post, excellent advice. This part in particular:
> When working with paid professionals, I think people tend to put them into one of two categories: doctors or general contractors.
> With doctors, if your oncologist says you have a lump that needs to come out, you pretty much do whatever they say. Youâre the patient; theyâre the expert leading the charge.
> With general contractors doing home renovations, most people intuitively know thatâs not the right approach. You stay on top of them, give clear instructions, try to understand the issues yourself, and assume you need to provide lots of direction and monitoring to get what you want done.
> Most peopleâentrepreneurs included, certainly myselfâput lawyers in the doctor category when they should really treat them more like general contractors.
Except... I think if one has enough time, it's better to treat everyone like general contractors. Doctors make mistakes, often. You can and should challenge them.