Forgot your password?
typodupeerror
AI

ChatGPT-Using Lawyer Cited Its Fake Witnesses and Police Testimony in Court (reuters.com) 98

Reuters reports: A defense lawyer appealing his client's murder conviction submitted a brief containing made-up police testimony and witnesses fabricated by OpenAI's ChatGPT, New Mexico's highest court said. The New Mexico Supreme Court on Wednesday fined the attorney, Stephen Aarons [for $5,000], and held him in contempt for failing to verify the accuracy of the court filing, which Aarons said he prepared with help from the AI program... and said they will refer him to an attorney disciplinary board for investigation.

Aarons in a statement to Reuters said he had used ChatGPT to summarize the trial proceedings when he agreed to take up the defendant's appeal last year, and did not understand the degree to which AI could "hallucinate" facts. "I am remorseful but hopeful that the disciplinary board takes into account it was an honest mistake," he said.... Dozens of lawyers have been sanctioned for filing briefs where AI made up case citations or misquoted the law. Aarons' filing appears to have gone further, containing fabricated witness testimony in a criminal appeal...

Aarons told the court at an August 21 hearing that he fed a computer-generated transcript and other case materials to ChatGPT, presuming it would generate "a bulletproof summary." The justices sounded incredulous that Aarons was not fully aware of how AI can make mistakes. "Counsel, do you watch the news? Do you listen to the radio? Do you read anything about what's going on in the world?" Justice C. Shannon Bacon said at the hearing. "Because the problem with lawyers relying on AI hallucinations is an above-the-fold story every single day."

"OpenAI did not immediately respond to a request for comment."

ChatGPT-Using Lawyer Cited Its Fake Witnesses and Police Testimony in Court

Comments Filter:
  • ... what makes anyone think there's a way to come up with a kill switch?
    • Chances are that power button on your phone is not really a power button but a soft suggestion. Kill switches are dead.
      • by Aighearach ( 97333 ) on Tuesday September 15, 2026 @03:26AM (#66338140)

        I don't know about any particular phone, but in many devices the long press power off is an old-fashioned kill switch that charges a capacitor, eventually reaching the voltage to turn off a transistor connected to a "power good" line on the power management IC. And inside the IC it's the same deal. So as long as the button works, it should be able to hard power off, even if the software is badly bugged.

        Actually numerous times I've seen people complaining their phone or tablet wouldn't turn off, when actually the software was just frozen and they didn't know about long press power off. "May I see that?" "Wow... that's amazing!" "To you, maybe..."

        • Not necessarily a capacitor; sometimes a little counter circuit, on the MCU but independent of the CPU and not controllable (or vetoable) by it. In either case, not prone to systemic failures of the higher-level logic.

          People overestimate the difficulty of adding low-level contingencies to high-level systems, and underestimate engineers' eagerness to do so. But this is the sort of thing that EEs worry about all the time.

        • Re: (Score:1, Funny)

          When I'm worried my phone doesn't turn itself off, I just use a small flyswatter.

          (And by small, I mean large.)

          (And by flyswatter, I mean hammer.)

          Works every time though!

          (But it's expensive! I really ought to try this long press thing one day.)

        • We bought some Dell towers last year where even the long press of the power button can apparently be overridden by software.

          I was updating several dozen of them using fwupdmgr (followed by a reboot), but one of them never completed the update after rebooting. I gave it 24+ hours, after which it was still showing a firmware update progress bar (at 0%). At that point, despite the on-screen warning, I decided it was time to power it off. Well, I ended up holding that button down for more than 60 seconds withou

          • If the long press isn't working, pull the plug. HTH, HAND.
            • Yup, that was actually was I was going to do next.

              Funny thing, though. After I eventually let go of the power button (it may have even been 2 minutes that I'd kept it depressed) with the intent of physically unplugging that box - the progress bar finally started moving, and the machine installed the various updates which it had staged.

        • No longer works on Pixels. Power button long press just brings up some kind of screen management app. There's no straightforward way to power off the device anymore at all. You have to go through a menu and a couple confirms.

          Terrible for police encounters.

          • by TWX ( 665546 )

            Terrible for police encounters.

            Not if you're the police.

          • My Pixel 9a doesn't really have a power button. Holding the top button brings up the AI assistant. You have to press and hold the top power and the volume up buttons, then, touch the option to power down or restart.

          • by sjames ( 1099 )

            If they really wanted to serve the user, it would understand "Hey Google, lockdown now" and would power off and require a PIN to restart.

        • by sjames ( 1099 )

          Long press doesn't do anything so low level anymore, it just signals the management processor in the radio that it should force the application processor off. The management processor keeps running.

          • You're not even bright enough to know the difference between some and all. You can't pass a basic "some foos are bars" logic test. All you do is spew BS and pretend you know more than everybody.

            • by sjames ( 1099 )

              Who wizzed in your wheaties?

              And I stand by what I said. Practically no phone that still works at all does it at the low level of a capacitor anymore.

      • by sjames ( 1099 )

        The power button on the phone really is just a suggestion. To actually power off the whole phone you have to remove the battery (if possible) or let it run down to zero.

    • Anecdotally, I've had pretty solid results with gemini and grok not hallucinating when it comes to the law. Unlike these lawyers, I ask the LLMs to provide direct links to the legal code and/or case briefs, and they're usually pretty spot on, usually only disagreeing with one another in subtle ways that, in my non-lawyer mind, seem to be the kind of thing that could go either way. After all, the reason you end up in court to begin with is to have the court interpret exactly these kinds of disagreements that

      • by Rei ( 128717 )

        I mean, isn't this obviously the solution?

        You can use AI in legal work, but you still bear responsibility for your filings being accurate. So by all means have AI search for you - it can save you a ton of time for finding obscure things - but actually click the links and read them yourself. And if you don't, it's no better than making any other factual claim before the court without having duly assessed its credibility.

      • by Tyr07 ( 8900565 )

        I don't get why people don't tell it that it has to review it's answer and summary against the source. I do that all the time. That's a SOP for any AI usage when I have it research, and provide me the sources.

        It pretty much never gives me a wrong answer as it self corrects before it tells me. My instruction set in there whenever I ask it something it goes, 'Let me verify directly instead of going off of the memory' and goes and checks, finds and makes corrections when needed and the results I get are correc

    • by Aighearach ( 97333 ) on Tuesday September 15, 2026 @03:10AM (#66338132)

      A "kill switch" is generally a hard-switch that removes power, it's not a soft-switch that the device can ignore or bug on.

      • by Teun ( 17872 )
        Once upon a time you could simulate this by taking the battery out...
      • by HiThere ( 15173 )

        No. Not "is generally", but rather "must be". If the software can decide to ignore it, it's not a kill switch. (OTOH, computers today don't tend to come with kill switches. Those would prevent remote wake-up commands from working.)

      • A "kill switch" is generally a hard-switch that removes power, it's not a soft-switch that the device can ignore or bug on.

        (AI Sidious) ”Good, goood. Now that we have the final design consideration for the power circuit..we shall begin the design.”

        (Meatsack Safety Manager) ”Hey uh, just curious..who changed the digital label on the kill switch to ‘Humans’..?”

    • ... what makes anyone think there's a way to come up with a kill switch?

      If a child was acting stupid and rambling on and on nonsensically last century, we used to remember what to do as adults. Spank the shit out of them and tell them to knock it off.

      And keep in mind I’m talking about what a teacher would do. Before the kid ever got home to a parent who would reinforce what the educator was teaching.

      When society can’t even remember how to discipline a child properly, the fuck makes you think AI will even NEED a kill switch?

      General intelligence and common sense i

      • by sjames ( 1099 )

        Back in the day when the "Board of Education" was an actual wooden board.

    • >... what makes anyone think there's a way to come up with a kill switch?

      What makes you think any of its Super Smart™ answers are anything more than hallucination wrapped in sophistry?
    • Unplug the wifi...

    • ...what makes anyone think there's a way to come up with a kill switch?

      There are myriad ways to deactivate a lawyer.

    • by HiThere ( 15173 )

      How long ago did this happen? The summary didn't list the date of the submission of the "testimony", but thing like this tend to take time to process, so I'm guessing that it happened well over a year ago. This make his claimed "honest mistake" due to ignorance believable. But it also negates your point.

      That said, they won't be able to stop hallucinations without breaking the AI models. That's a mistake that people make all the time. What they need to do is prioritize having the AI determine how certai

      • by pla ( 258480 )
        It's an "honest mistake" by the AI; it's gross negligence by the lawyer, for completely failing to do his actual job and confirm what his research tools spit out.

        More to the point - "Hallucination" is an awful description and doesn't mean what most people think it does. If you force a human to answer a question regardless of certainty, you'd get the exact same behavior. High-confidence answers are easy; small medium-confidence details are likely to be in the ballpark but maybe a bit off; at the bottom of
    • by Lehk228 ( 705449 )
      all that is needed is a hard prohibition on integrated power generation on the same site as AI data centers, even having power for the center is ok but it must be outside the perimeter of the building such that the transformers can be shot out if necessary in an emergency
  • by Mr. Dollar Ton ( 5495648 ) on Tuesday September 15, 2026 @01:54AM (#66338110)

    Most people hit by the "AI revolution" that I know have degraded rapidly by the daily reliance on their best friend.

    And it will get a lot worse. But on the positive side, the AI is ready, once I asked it something like "mmhhh ggghhh rrhhhrmmhhr?"

    It immediately replied, "what would you like, a greasy burger, an energy drink or a cool tune to rave to, give me one number and I'll tell you how to get one".

    We'll be getting there fairly soon.

    • Re: (Score:2, Interesting)

      by Powercntrl ( 458442 )

      Most of us here should be old enough to remember when people tried to do serious work under Microsoft Windows in the MS-DOS era, and it crashed if you looked at it funny. Another amusing "it's never going to work" technology story was the development of lithium ion batteries. Early attempts to design a working cell kept going up in flames - literally. [thermofisher.com]

      It was even one of the old running gags in Disney's Carousel of Progress that new technologies were inherently buggy.
      "But we do have television - when it wor

      • by kertaamo ( 16100 ) on Tuesday September 15, 2026 @04:13AM (#66338162)

        I'm old enough to do a lot of work using MS-DOS. I was on a team that developed a CAD system for schematic capture and printed circuit board layout. For years I used MS-DOS for software development. I don't recall anyone saying "it's never going to work". Even if MS-DOS was a simplistic and crude platform. Mind you I did have two Microsoft Surface Pro's whose batteries expanded enough to push the displays off the machine threatening to go on fire.

      • Yeah there's a bit of that going on. But Microsoft didn't claim it was going to make most work redundant and suck up all investment from something actually useful.

        AI has some irresponsible marketing bullshit that's got a lot of folks laid off and leaving the ones behind to work harder.

        MS Windows created lots of IT support roles.

        • It is also creating jobs. No different than what a spreadsheet did to acocunting, what the car did to the buggy whip industry, what computers did to people who did that work by hand, what the internet did to printed research material, what smartphones did to dozens of tools it replaced, electric lighting did to lamps, the printing press did to hand copying, digital photography did to dark rooms, CAD did to drafting.

          People lost their jobs at first, then new jobs using and leveraging that tech replaced them.

          • by TWX ( 665546 )

            Except that in basically all of the examples that you cite, the transition was from one proven technology slowly into another technology that was at least approaching some degree of maturity by the time it it mass-market adoption.

            In most of those examples. early adopters were nearly all ultrawealthy who were using it for themselves. Those who weren't incredibly rich were technology-enthusiasts of some fashion or another. Those groups initially worked-out most of the showstopper-problems with the technolog

            • I think you are framing your argument from the luxury of the modern era. Early adopters didn’t finish working out the serious problems with cars, machinery, or electricity before widespread adoption. Safety standards and regulation often developed in response to harms that were already happening. In fact, this is happening today with AI. Tech leaders are calling for better standards and even regulation to reduce the harms AI is causing.

              Some of us here were lucky enough to live in a world without leade

            • by sjames ( 1099 )

              Exactly. For example, spreadsheets were often used day to day but the accountants kept the actual books as books and had the final word. That changed over time, only after spreadsheets proved out (and got the bugs out).

              Also, humans stayed in the loop. Sure there were occasional funnies like a medical database claiming a few male patients had a history of pregnancy, but humans were in the loop and cited those as obvious errors. There were still backup paper records to fall back on.

          • It is destroying far more jobs than it is creating, so no it's not creating jobs, because when we say that what we mean is that it's creating NET jobs. So just no.

            Also, even if that were not true, it's doing it at a rate that will kill people, if it hasn't already. Therefore celebrating it or even making excuses is ghoulish at best.

            • Like industrialization? I'm not saying it's for the best. It is impossible to know if technological progress is for the best. I'm saying it is no different than any other kind of technological progress and we have to make a choice. Protect jobs at the cost of progress or accept that this is the cost of progress.

              Sure I'd love for there to be a generation of 20 something hand crafters carving the scroll work on my shotguns, but the market has decided that engraving machines are cheaper and good enough. So tha

          • Yours is the anti luddite argument.

            How are those innovations comparable? They all improved or replaced something.

            What does AI, a fallacious term, improve or replace ? Intelligence ? Well it's not Intelligence, for the simple reason it doesn't know when its wrong. It has no concept of true or false, at best an approximation of probability provided it has content or user feedback.

            Search, summary and content creation it makes easier and quicker, but they are only part of an improvement, it fails or is at leas

      • It does work, people are getting more stupid, enormous resources are misallocated in a terribly inefficient manner, money and power are concentrated, etc. and it is a process that is basically irreversible.

        That it produces mountains of shit as an outcome is a feature, not a bug.

      • I have a few colleagues who, if you take their conversation too serious, became millionaires by doing clever stock trades. For some reason my youtube feed is filled with people telling me that the end of the bubble is near these days, so I joined their conversation. It bought back memories about the high days of facebook. It was revolutionary back then. (Web 2.0 !) But... it was burning money. How is this webpage ever going to make money. That was a big thing when it went to the stock market. It was a bad i
      • by sjames ( 1099 )

        But a lot of people when they hear someone say "it's nowhere near ready for prime time" will claim the person is saying "It'll never work".

      • They threw that version of Windows away and wrote another one. We were correct, that form of Windows was never going to work right.

        In fact, even though they rewrote it, Windows is still shit and unfit for purpose. Wait, what was your point again?

    • Most people I know haven't had the degradation issue, although I have seen it in some people. I find a bigger problem is people believing bad data/ information provided by LLM's over Subject Matter Experts (SME) who are providing information from the field.

      They also react very negatively to you when you show them that they were given bad information.
  • by Fly Swatter ( 30498 ) on Tuesday September 15, 2026 @02:54AM (#66338120) Homepage
    Ignorance of a broken tool combined with zero review of your own court filings should be more than discipline, it should be disbarment.

    - don't practice law without a brain
    • Exactly.
      $5000? It's like fining Apple one million dollars. Pocket change.
      Immediate disbarment is the bare minimum. If he is too stupid/ignorant to check his own submission, he has no right to be a lawyer. Not to mention courts will be even more swamped by the AI slop than they already are with proper submissions. Soon even judges will have to use AI and then what will be the point of a justice system?
    • by HiThere ( 15173 )

      It sounds as if he was referred to the bar association evaluation committee. IIUC those are he folks that can do the disbarment. A court can only say "you can't be a lawyer before us anymore", but he'd still be fine in a different court.

  • Contempt of Court (Score:5, Insightful)

    by Errol backfiring ( 1280012 ) on Tuesday September 15, 2026 @04:17AM (#66338166) Journal
    There is no worse contempt of court possible than to throw an LLM-generated word salad over the court wall. This "lawyer" should never be able to work in justice again.
    • by evanh ( 627108 )

      Yup, and I think that's exactly how the judge has seen it too. And I'd expect the investigation to conclude the lawyer is a lying little shit.

    • by Guignol ( 159087 )
      Sure, yeah, but, the lawyer could also just be wrong or lying, in the end he presents his side of the story which is to be as convincing as possible with wildly open to interpretation word salads, you know, human talking to human.
      I find it terrible that he would use AI to do that kind of things instead of doing it himself, certainly.
      But it would be orders of magnitude worse if he used AI to produce a highly precise, very structured text, with 0 margin for interpretation and for which any 'hallucination' m
  • I mean really ... I know it's absurdly easy to practice law in the US, but really ... come on.

  • At least pay to use a frontier model for the highest quality nonsense.

  • "Human lawyer using AI tool" clearly is not the way to go these days, instead the lawyer should have been replaced by an AI agent (owned by some big corporation), because we all know when such AI agents do wrong, nobody gets fined, and the event would even be good for some marketing press releases.
  • How many cases of AI hallucination are going uncaught?

    We know how prosecutors and law enforcement can get tunnel vision when dialing in on someone they want to be guilty.

    It would be very easy for AI hallucinated details that fit the narrative to pass muster or go unnoticed all the way through to a conviction, at which point it's too late.

  • Sure sounds like Perjury or at least Subornation of Perjury to me. When a lawyer's "client" does that, they do jail time.
    How about when an officer of the court does it, shouldn't that be worse? Yet more douible-standard "justice".

    • by TWX ( 665546 )

      The only place I see Al in the legal profession making sense is to do basic caselaw research, where it can cite the sources and the references upstream and downstream to those sources. The lawyer should be responsible to review the claimed caselaw citations to determine the merit of their support for the argument that the lawyer is making, and to pick the particular citations to use that are the strongest for the particular case.

      Likewise the same sort of search but inverted, for one's opponent's most likel

    • I don't know about all that but it seems to me like the lawyer's client has a fraud case against him. The lawyer didn't actually do the job.

  • by CommunityMember ( 6662188 ) on Tuesday September 15, 2026 @09:45AM (#66338386)
    There are (at least) a couple of firms who claim they are a "Fiduciary-Grade AI", trained only on the appropriate laws and cases (I also presume those solutions are not at all cheap). ChatGPT does not make those claim, and any lawyer who uses it is more likely than not to have chosen poorly (and subject to sanctions and discipline if they did not re-validated each and every claim made in their filings).
  • The AI thought that laws in New America allowed this.

  • by reanjr ( 588767 ) on Tuesday September 15, 2026 @10:28AM (#66338450) Homepage

    Judges need to start applying eye watering fines so the message gets through. They have no problem sending a message to drug dealers and murderers. Now step up, do your job, and send a message to your buddies at work.

    • This is exactly what I was going to post: Add more zeros to the fine. These lazy fuckwits need to be taught a lesson and $5000 is only a few hours for them.
  • We think AI is intelligent because it appears to communicate. Bees communicate.

    AI is about as dumb (not smart) as bees. They can communicate but they do not understand what they are saying. They have a massive hive to repeatedly do dumb tasks till they give you what you asked for. But you asking is what triggered the action. Your question is the only intelligence involved, what they did was the dumb busy bee work.

    LLM are not intelligent, they are artificially stupid.

  • ... did not understand the degree to which AI could "hallucinate" facts.

    As the judges implied, I'm pretty sure EVERY practising lawyer has heard about AI hallucinations biting lawyers on the ass. This guy is either lying or hopelessly out of touch.

    Either way, he should be disbarred because he is either too dishonest and corrupt, or too stupid, to practise law.

  • LLMs used to hallucinate a lot. It was fairly common. However, when working with supporting documents that the LLMs can reference as context, I haven't seen an LLM hallucinate in months. The things have gotten reliable enough that it's almost more plausible to me that someone's making stuff up, and using that as input for the LLM, and then blaming the LLM for the fabrications.

    Over-confident? Still a problem. Sycophant? Still a problem. Able to be confused with garbage input? Still a problem. Making stuff up

  • Everyone knows AI can make stuff up, my grandmothers, one is 86, the other is 94, both know that AI makes stuff up. If a lawyer can't be bothered to read what they submit, why give them an easy excuse / pass? Forget about AI, frankly lawyers mess too much up anyway, and it's unacceptable. I've seen contract reviews, policy reviews, license reviews, where they'll gloss over statements, then act like they didn't when you point that out.

    I just finished a policy review for a company I won't name. In two o
  • The lawyer in question here claimed he didn't know that AI can hallucinate. What better way to have the message spread than make an example out of him and any other lawyers doing the same thing? Humans learn best on extreme examples, so disbar the lawyer and make them pay for a retrial including a replacement attorney of their client's choice (which I'm sure will be the top firm in the country), and such a story will very quickly spread through the attorney community, therefore making all lawyers aware of t

I find you lack of faith in the forth dithturbing. - Darse ("Darth") Vader

Working...