This sounds paranoid but: Trust no one

I really like writing this blog and I've been writing it on and off for about 14 years.

That's right - 14 YEARS! 

Some topics are better than others but what has really been helpful is this new AI set up they have going.  

Specifically, Chatgpt has come in handy helping with identifying topic, cases, and statutes relevant to the topics I'm writing about.

The problem with these AI search engines like ChatGPT is that you can't really depend on them for accuracy when researching law or things of a legal nature.  Sometimes, it even creates cases out of thin air.

Wait, what?!  

You mean to tell me that AI search engines spit out caselaw that doesn't even exist or out and out misleads researchers?!

Called Hallucinations and yes, that's exactly what I mean.  

Because LLMs (i.e. "Large Language Models" which is the official classification for AI platforms) create hallucinations with reckless abandon, you have to be extra careful when using anything you find on an AI search engine.

For example, do you remember Mata v. Avianca, Inc., No. 22-CV-1461 (PKC), 2023 WL 4114965 (S.D.N.Y. June 22, 2023)?  In Mata, lawyers submitted a brief with six case citations (and quotes) that did not exist; they were generated/made up by ChatGPT.  The court imposed a $5,000 fine on the lawyers and their firm (which is peanuts).  The judge found they acted in bad faith, including conscious avoidance of verifying the non-existence when there were signs the cases might be fake.  

Or how about United States v. Cohen, 724 F. Supp. 3d 251 (S.D.N.Y. 2024).  In that case Michael Cohen admitted he had given his lawyer fake citations generated by Google Bard. The lawyer then submitted them.  The court called the inclusion of the fake cases “embarrassing and certainly negligent, perhaps even grossly negligent.” But the judge determined there was no intention to deceive, i.e., not in bad faith, so no sanctions were imposed.

Other cases where attorneys used holographic cases (ie. cases generated by AI search engines that don't exist) include:

Garner v. Kadince, 20250188-CA (Utah Ct. App., filed May 22, 2025): Their petition included case-citations that were found to be fabricated or “hallucinated” by AI (ChatGPT). One example: Royer v. Nelson, 2007 UT App 74, 156 P.3d 789, a case that does not exist in any legal database except in AI outputs.

Wadsworth v. Walmart (D. Wyo. 2025): Three lawyers cited fake cases that were AI-generated in a personal injury lawsuit against Walmart. 

Alabama Prison Litigation (2025): Butler Snow firm cited entirely fabricated Eighth Amendment cases.

My personal opinion is that any attorney that doesn't Shepardize ALL cases cited in legal documents should be, in the very least, suspended for at least a 6 months.  They should also be required to pass the Multistate Professional Responsibility Exam (aka MPRE) before being again allowed to practice law. 

But how does this all affect me, the legal blogger?

Over the last year, I've blogged about a whole range of legal topics.  In a few instances, I've found, through ChatGPT and other AI search engines, not a few discrepancies.

For example, recently I wrote a blog about tailgating and after running a search in ChatGPT for cases related thereto, I got a number of cases back presumably on the subject.

One case in particular, Shuck v. Means, was particularly difficult to locate on the Internet and I asked ChatGPT:

 is Shuck v. Means, 302 P.2d 217 (Okla. 1956) a real case and does it have anything to do with tailgating

To which ChatGPT responded:

You're in for a quick clarification—“Shuck v. Means, 302 P.2d 217 (Okla. 1956)” does not appear to be a real case, at least not under that name or citation. 

A few paragraphs later, I ran into a similar problem when after a query ChatGPT returned a case (in this instance, Baker v. Rushing).  After an extensive search on the Internet of not finding that case, I asked ChatGPT:

is Baker v. Rushing, 104 S.E.2d 165 (N.C. 1958) correct in this format

To which Chatpgt responded: 

Yes — that citation format is correct. Here’s the breakdown:

• Case Name: Baker v. Rushing

• Reporter Volume: 104

• Reporter Abbreviation: S.E.2d (South Eastern Reporter, Second Series)

• Page Number: 165

• Court and Year: (N.C. 1958) → North Carolina Supreme Court, 1958

So it should appear as:

📖 Baker v. Rushing, 104 S.E.2d 165 (N.C. 1958).

That is a valid, properly formatted Bluebook citation.

Now, to the untrained eye, that would look like this is a real/actual case and a rookie researcher would stop there.  I mean, it says it's the right Bluebook format - so that should mean it's a real case.  However, after I couldn't find Baker v. Rushing, 104 S.E.2d 165 (N.C. 1958), I went back to ChatGPT and again asked:

is Baker v. Rushing, 104 S.E.2d 165 (N.C. 1958) a real case

 To which Chatgpt responded:

Great question — “Baker v. Rushing, 104 S.E.2d 165 (N.C. 1958)” does not appear to be a valid or real case citation, at least based on available legal databases and authoritative sources.

...which is annoying - and definitely misleading - given that ChatGPT first said the case was in the correct Bluebook citation format (which was technically true) but then acknowledged that it screwed up and returned a case that was neither valid or real.

Soooooo aggravating! 

Not wanting to drop the issue, I again asked ChatGPT:

if Baker v. Rushing, 104 S.E.2d 165 (N.C. 1958) is not a valid or real case, why did you return it in my initial query 

To which ChatGPT responded:

That’s a very fair question — and you’re right to call this out.

Here’s what likely happened:

When you first asked about tailgating-related cases, I tried to retrieve examples of real case law, but I sometimes draw from patterns in legal writing rather than verified court databases. “Baker v. Rushing, 104 S.E.2d 165 (N.C. 1958)” looks like a perfectly plausible citation (correct reporter, correct jurisdiction, correct year), but after checking, there’s no record of such a case.

This is what’s known as a hallucination (or “holographic case”), where I construct something that looks real based on legal citation formats and similar case names — but it doesn’t actually exist.

The main reasons this happens are:

• Predictive nature of my training: I was trained to generate text that follows patterns, including how case citations are written, but I don’t have perfect access to every case law database.

• Similarity to real cases: There are real cases with names like Baker v. [Someone] and reporters like S.E.2d from 1958, so it can produce something that seems legitimate.

• Overconfidence in early responses: Without doing a real-time case law search (like we just did for your follow-ups), I might give examples that aren’t verified.

This tends to happen more and more and as I use AI to help with blogging, I now ask whether particular cases returned are real before including them in my blog.

But this got me to thinking and I asked ChatGPT:  

why does chatgpt return cases in a search that don't exist?

To which ChatGPT responded:

That’s a great question — and it’s one of the most important issues when using AI for legal research.

ChatGPT (and other LLMs) sometimes return cases that don’t exist because of the way large language models work:

• Pattern Matching, Not Databases:
  I don’t have a built-in legal case database. Instead, I was trained on lots of legal writing and I generate responses by predicting what looks like a plausible answer. If you ask for a case citation, I try to produce one that fits the pattern of a real case — correct reporter, volume, page — even if it’s made up.

• Pressure to Be Helpful:
  If you ask for “a case where X happened,” I try to fulfill that request. If I don’t have a known case in memory, I may invent one that sounds right rather than say “I don’t know,” especially if your prompt suggests that such a case must exist.

• Legal Citations Are Predictable:
  Because case citations follow a rigid format (e.g., Smith v. Jones, 123 F.3d 456 (9th Cir. 2020)), I can generate something that looks very convincing — even to experienced lawyers. But if you check the reporter, the case might not exist.

• Contextual Merging (Hallucination):
  Sometimes I blend real cases. For example, I might take facts from Case A and a holding from Case B and give it a citation that matches neither. This is not intentional — it’s a statistical artifact.

This problem is exactly what led to the high-profile Mata v. Avianca (S.D.N.Y. 2023) case — where lawyers used ChatGPT to write a brief, it cited six fake cases, and the court sanctioned them.

✅ Best Practice When Using AI for Legal Research:

• Treat all AI-supplied case names and citations as leads, not authorities.

• Verify every case in a trusted source (Westlaw, Lexis, Fastcase, Casetext, etc.) before using it.

• If you can’t find it in an official reporter or database, assume it doesn’t exist.

And there you have it - even ChatGPT knows it lies to/misleads people.  Well, maybe not out and out lie but returns results that are not accurate or existent.

 

For this reason, 

• always verify results, 
• always recheck results, and 
• never trust that the bot you're talking to is telling you the truth

...because what does an artificial intelligence know about "truth" when any/all information is true to an AI bot?

 

 

An Ounce of Prevention

Have you ever had those moments when you thought to yourself - dang, what I should have said was....  Or maybe you did something and later wished you hadn't.

Maybe like when your boss said something condescending to you and at the time the only thing you thought to say was, "That's a good one, boss" then later you thought, what I should have said was "I came here to do good work—not absorb bad management!"

Or maybe when you asked your wife if she minds that you go golfing with your buddies and she says that's "fine" - so you go and when you get back she's pissed off seven ways from Sunday but you don't know why.

Stuff like that.

This got me to thinking the other day when a friend of mine related an interaction he had with a local "Karen" in his neighborhood.  Seems said "Karen" had been riding his backside for the last few months because his Christmas light were still up.

Thing is, there are any number of reasons "Karen's" go off on people but it got me thinking of ways to avoid confrontational neighborhood "Karens."  So, I create a list of ways to keep your neighborhood Susan's or Jennifer's or Laurie's from turning into a Karen.

1. Make your property “boringly compliant”

Most neighbor conflicts come down to things that look like violations—even if they aren’t.  “Make your property ‘boringly compliant’” means keeping everything so clearly within local rules and neighborhood norms that there’s nothing for anyone to latch onto or argue about.  For instance:

• Keep lawns, weeds, trash bins, and parking tidy
• Check your city’s basic ordinances (noise hours, fence height, parking rules)
• If something looks borderline, assume someone will notice

This isn’t about surrender—it’s about removing easy ammunition.

Related Authority:

Village of Belle Terre v. Boraas, 416 U.S. 1 (1974) reinforced the authority of municipalities to establish and enforce land-use regulations. NOTE: the standards come from the ordinance, not the loudest neighbor.

 2. Control the first impression

People are far less likely to escalate if they see you as human.  “Control the first impression” means making sure neighbors see you as a normal, reasonable, low-drama person early on—because people tend to interpret everything you do later through that initial lens.  This can be accomplished by:

• A simple wave or quick introduction early on goes a long way
• You don’t need to be friends—just recognizable and polite
• Avoid coming off defensive or standoffish right out of the gate

It’s harder to complain about “that guy” if you’ve already humanized yourself.

Related Authority:

Prah v. Maretti, 108 Wis. 2d 223, 321 N.W.2d 182 (1982) recognized that neighboring landowners owe one another duties of reasonable use and that disputes between neighbors should be evaluated in light of what is reasonable under the circumstances.

3. Don’t feed the escalation loop

“Don’t feed the escalation loop” means resisting the urge to match someone’s intensity when they come at you with complaints, attitude, or accusations. The more emotion, defensiveness, or sarcasm you add, the more the interaction tends to spiral.  If someone nitpicks or complains:

• Stay calm, brief, and neutral
• Don’t argue on the sidewalk—it turns into theater
• If they’re wrong, you don’t need to prove it immediately

Escalation is oxygen. Starve it.

Related Authority:

In Rattigan v. Wile, 445 Mass. 850, 841 N.E.2d 680 (2006), a disagreement between neighbors escalated into a deliberate campaign of retaliation that ultimately resulted in liability for private nuisance. The legal lesson is simple: once a dispute becomes about "getting even" instead of solving the problem, every retaliatory act increases the likelihood that a judge (not just the neighbors) will become involved.

4. Use quiet documentation

Keep a calm, private record of what’s happening instead of reacting in the moment or getting pulled into arguments.  The point isn’t to build a case theatrically; it’s to create clarity and protection for yourself if things escalate.  If someone is persistently difficult:

• Keep records (dates, what happened, photos if relevant)
• Save texts or written communication
• Know your HOA or city complaint process before you need it

This flips the power dynamic without confrontation.

Related Authority:

Courts recognize that memories fade, stories change, and disputes become harder to resolve as time passes. Contemporaneous records—such as photographs, messages, logs, and written observations—can be powerful evidence because they capture events closer to the time they occurred. See United States v. Orr, 68 F.3d 1247, 1251–52 (10th Cir. 1995).

5. Choose your battles (strategically)

Recognize that not every annoyance, comment, or minor rule-bending is worth your time, energy, or reputation to confront.  The strategic part is deciding what actually has long-term impact: safety issues, repeated harassment, property damage, or whether formal violations are worth addressing.  So:

• Minor annoyance → let it go
• Pattern of harassment or false complaints → address formally
• Safety or legal issues → act immediately

Winning every small fight can lose you the bigger war.  What this means is that a person who complains about everything risks looking unreasonable, but a person who ignores a pattern of escalating conduct may lose the ability to prove the seriousness of the problem later. The strategic choice is not "never complain"; it is reserve formal action for things that have legal or practical consequences.

Related Authority:

The law recognizes that neighbors will inevitably annoy each other from time to time. Courts generally distinguish between ordinary inconveniences and conduct that creates a substantial interference with property rights. See Armstrong v. Francis Corp., 20 N.J. 320, 120 A.2d 4 (1956). 

6. Set calm boundaries

Clearly define what behavior you will and won’t engage with, but you do it without anger, lectures, or emotional escalation.  In practice, this looks like short, repeatable statements.  You stay consistent—calmly reinforcing the same limit if needed.  If someone crosses a line:

• “I’m happy to follow city rules, but I’m not going to engage beyond that.”
• Repeat as needed, without adding emotion

Consistency beats clever comebacks.

Related Authority:

In Rowan v. United States Post Office Department, 397 U.S. 728 (1970), the Supreme Court upheld a homeowner's right to stop unwanted communications from entering the home. The practical lesson is simple: state the boundary clearly, avoid unnecessary debate, and do not turn every disagreement into a confrontation. 

7. Build a buffer (social proof)

Quietly create a network of normal, neutral, or positive neighbor relationships so you’re not isolated or easy to target if conflict ever pops up.  This is important because neighborhood disputes often aren’t just about facts; they’re about perception and who seems credible.  If you have even one or two neutral-to-positive neighbors:

• Casual friendliness creates a “baseline reputation”
• People are less likely to side against someone who seems reasonable

It’s subtle, but powerful.

Related Authority:

Credibility, reputation, and perception matter when evaluating disputed claims. In Bose Corp. v. Consumers Union of United States, Inc., 466 U.S. 485 (1984), the SCOTUS discussed the importance of reputation and the difficulty of determining truth when competing accounts exist.

8. Worst-case: disengage + formal channels

Recognize when direct, informal interaction has stopped being productive and switching to a more structured, low-contact approach is a better approach.  If someone is truly relentless:

• Stop direct interaction
• Communicate only in writing if necessary
• Use HOA, mediation, or local code enforcement appropriately

At that point, it’s not about winning socially—it’s about protecting your time and sanity.

Related Authority:

Courts recognize that when disputes become persistent and informal efforts fail, the solution is often a structured remedy - not more arguing. See Boomer v. Atlantic Cement Co., 26 N.Y.2d 219, 309 N.Y.S.2d 312, 257 N.E.2d 870 (1970).  Also, in Rondeau v. Mosinee Paper Corp., 422 U.S. 49 (1975) the SCOTUS discussed the availability of equitable relief when legal rights require structured judicial intervention.

The thing is, most people want to be reasonable.  Rare is the case that people flip out over kids playing in the street, or neighbors having a block party, or someone installing a pool in their backyard.

While most people want to be reasonable, some don't.  In such cases, keep a record, take notes, make friends where you can and, maybe, just maybe, buy one of those REALLY loud 1,000 watt speakers to blast Mozart's Queen of the Night aria at at all hours of the day/night and engage in some serious psychological warfare.

 

I mean, if Karen wants crazy, I suggest going lunatic crazy.  

Go all out.  

Make a scene.  

Give as good as you get.  

It may be counterintuitive, but sometimes crazy is good for the soul.

 

 

Word of the Month for March 2026: Fake News

I know you've heard the term Fake News.  It was Trump's rallying cry for several years and he says it whenever any news organization reports something he really (or marginally) doesn't like.

That's not what we're talking about here.  

What I'm looking at here are those cases where the plaintiff (or defendant) is being castigated by the press notwithstanding the fact that the plaintiff (or defendant) in a real case actually has a legitimate claim but the press is pushing the narrative against that of the defendant (or plaintiff).

So, let's dispense with pleasantries and define Fake News in this context.

FAKE NEWS constitutes a subset of intentionally or negligently disseminated pseudo-journalistic content, characterized by the deliberate fabrication, distortion, or strategic omission of verifiable empirical data, disseminated via digital or traditional media channels under the guise of legitimate reportage, with the intent or foreseeable consequence of influencing cognitive biases, socio-political perceptions, or behavioral responses within target populations, thereby undermining the epistemological integrity of information ecosystems and the democratic discourse reliant upon them.

Ummmmmm, sorry about that.  My inner-journalist generator took over for a second.  Let's try that again:

FAKE NEWS is false or misleading information presented as news, typically to influence public opinion, generate clicks, or cause confusion.

OK, that's better - a story used to generate or cause confusing information - sounds about right.

So, how might this apply in real life?  Remember a few years ago, the media flipped out on a story about a bunch of kids and some Indians marching in D.C.?

In January 2019, Nick Sandmann, a Covington Catholic High School student, was filmed wearing a MAGA hat while standing face-to-face with Native American elder Nathan Phillips during the March for Life in Washington, D.C. A short video clip went viral, with media outlets and social media users accusing Sandmann of mocking or intimidating Phillips.

Later, fuller video footage showed that Sandmann was standing silently and that Phillips had approached Sandmann, while a different group (Black Hebrew Israelites) was yelling insults at the students.

Sandmann filed defamation lawsuits against major media outlets (CNN, Washington Post, NBC, etc.), claiming the coverage was "false and damaged his reputation." Several settlements were reached and Sandmann’s case became a key example in debates over media bias, viral outrage, and defamation.

See, that's what we're talking about.  Instead of reporting the actual events, news media is going around reporting lies and misstatements - instead of what is actually happening.

Another example?  

Recently, a Minnesota family, the Moedings, were sued by their neighbors, the Ramoses, over the location of their basketball hoop and the subsequent retrieval of balls that bounced into the Ramos' yard.  The lawsuit, which included claims of trespass and a request for an injunction to prevent the hoop's use, was ultimately dismissed by a judge. 

On first read (and media makes it look like), this sounds like a petty case where an older couple is trying to keep kids from using their (clearly) portable basketball hoop.  However, this case had little to do with basketball as much as the Ramoses were trying to keep the kids from trespassing on their land.  

See, the kids would play on their side of the property line but, as kids are wont to do, the ball would get away from them, bounce/roll to the Ramoses property and the kids would run across the property line trampling grass, flowers, whatever else is in their way to get their ball.

From a property owner standpoint, that would piss me off and if you know anything about getting old and how little things really annoy, it really rakes the nerves.  Anyway, the court eventually dismissed the case leaving the Ramoses little recourse but to sit and fume.  

While an easy remedy would be to have the Moedings install a net/fence behind the basketball hoop to keep the balls on their side of the property line - but that would be admitting fault. 

Some other cases found in the annals of history include: 

Richard Jewell - Atlanta Olympic Bombing (1996)
Plaintiff: Richard Jewell (security guard wrongly suspected).
Claim: Defamation against media outlets (NBC, Atlanta Journal-Constitution, CNN).
Media narrative: Initially pushed the FBI and law enforcement leaks implying Jewell was the bomber.
Reality: Jewell was innocent and later cleared, but the press framed him as seeking attention.
Outcome: Jewell settled with NBC and CNN; won undisclosed settlements.
Press alignment with defendant narrative: The FBI’s suspicion narrative was amplified despite thin evidence, harming Jewell’s life while media framed him as the villain.

Gibson’s Bakery v. Oberlin College (2019) 
Plaintiff: Gibson’s Bakery (family bakery in Ohio).
Claim: Defamation and tortious interference.
Media narrative: Many outlets leaned toward framing the bakery as racially discriminatory after an incident with Black students.
Reality: The students had shoplifted; evidence showed no racial discrimination.
Outcome: Jury awarded $44 million (later reduced to about $25 million) to Gibson’s Bakery.
Press alignment with defendant narrative: Several academic and progressive outlets pushed Oberlin’s framing that the bakery was targeting students based on race, despite facts supporting Gibson’s position.

Rolling Stone “A Rape on Campus” Defamation (2016)
Plaintiff: Nicole Eramo (UVA dean) and fraternity members.
Claim: Defamation over false rape allegations reported in Rolling Stone.
Media narrative: Rolling Stone and supporting outlets pushed the narrative of institutional coverup and fraternity violence.
Reality: The story was fabricated; no evidence supported the claims.
Outcome: Eramo won $3 million; fraternity settled for $1.65 million.
Press alignment with defendant narrative: Rolling Stone and sympathetic media initially defended the article despite growing evidence of its falsehood, framing plaintiffs as silencing victims.

Duke Lacrosse False Allegation (2006–2007)
Plaintiff: Three lacrosse players—Reade Seligmann, Collin Finnerty, and David Evans
Claim: Defamation of false rape allegations reported by CBS News.
Media narrative: In March 2006, Crystal Mangum, an exotic dancer hired to perform at a Duke University lacrosse team party, was raped by the three boys.
Reality: The story was fabricated and in December 2024, Mangum publicly admitted on a podcast that she had fabricated the rape allegations.
Outcome:

1. Prosecutor Mike Nifong aggressively pursued the case but withheld exculpatory DNA evidence.  
2. On April 11, 2007, North Carolina Attorney General Roy Cooper dropped all charges.  
3. The team’s season was canceled. 
4. Coach Mike Pressler was fired.
5. Nifong was later disbarred and convicted for misconduct.  
6. The players sued Duke and the city, reaching settlements. 
7. Mangum never faced charges for the false claims.  
8. Nifong served a day in jail for contempt.

Press alignment with defendant narrative: The charges triggered intense media coverage and national outrage, with debates over race, class, and campus culture fueling the story.

I guess the moral to this story is: Take everything big media says with a grain of salt realizing that the press is in the game to make money.  

...and, as it appears, the only way the Press can make money is to make-up stuff to stir up the blood of their victims (in this case, victim=anyone willing to listen).

Word of the Month for February 2026: AI

 

OK, OK, so "AI" is not, per se, a word so much as it is an acronym for "Artificial Intelligence."

Great and with that out of the way, what is AI (or artificial intelligence)? 

An overly complex definition of AI is: Artificial Intelligence (AI) is a multidisciplinary domain within computer science and cognitive science that involves the design, development, and analysis of computational systems capable of performing tasks traditionally requiring human cognitive processes such as perception, reasoning, learning, decision-making, and natural language understanding. It encompasses the creation of algorithms and models that enable machines to acquire representations of their environment, generalize from data, adapt to new information, and exhibit goal-directed behavior under varying conditions of uncertainty. AI draws on subfields including machine learning, knowledge representation, heuristic search, and robotics, leveraging statistical methods, neural architectures, and symbolic reasoning to enable autonomous or semi-autonomous systems to optimize actions in complex, dynamic environments while adhering to constraints defined by computational, ethical, and social considerations.

Got all that?

In more simplistic terms, AI is basically a fancy robot brain that tries to fake being smart so you don’t have to be.

That better?

Essentially, Artificial Intelligence is like building a mechanical apprentice that learns by watching, listening, and practicing, just as a human would, so it can help us carry out tasks.

For example:  Imagine teaching a child to sort laundry by colors: you show them examples, correct mistakes, and eventually, they learn to do it on their own.  AI works similarly, but instead of a child, it’s a computer system that learns from examples, patterns, and feedback so it can make decisions, recognize speech, translate languages, or drive a car.

It’s not truly “thinking” like a human, but it mimics parts of human learning and decision-making to help us do things faster, more consistently, and often on a much larger scale.

Still unclear how it works in "real" life?  

Say you're looking to write draft a professional resume for a sales professional (selling cars) and where you only have a few key skills that might be useful and you've worked at McDonald's slinging burgers for the last few years.

AI can crank a really nice one page resume, based on those parameters, for you.  Of course, I'd suggest you take the time to make small edits - but it will look sharp.

Maybe the resume you submitted above landed you an interview in front of 21 people.  While only 4 people asked you questions, you still need to send a thank-you letter to all 21 people.  I've done this and it took me 4 days to make each one a little different but relatable using notes I took during the interview(s).

AI can crank out those 21 unique and professional letters just based on their titles alone and do it in under 2 minutes flat and make you look like a superstar.

Maybe you're a lawyer and you need help with your lawyer stuff.  How might AI help you?

• Streamlined Legal Research:  AI can quickly analyze vast amounts of legal data, identify relevant precedents, and suggest potential arguments, saving lawyers significant time and effort. 
• Automated Contract Review:  AI tools can scan contracts for key clauses, potential risks, and inconsistencies, accelerating the review process and improving accuracy.
• Enhanced eDiscovery:  AI can help manage and analyze large volumes of data during litigation, identifying relevant information more efficiently and reducing costs associated with discovery. 
• Improved Risk Assessment:  AI-powered tools can analyze historical case data and predict potential outcomes, enabling lawyers to better advise clients and mitigate risks.
• Drafting Legal Documents:  AI can assist in drafting initial versions of motions, briefs, contracts, and other legal documents, saving time and improving consistency.

Heck, I used to consult for Anylaw.com where AI is a huge part of it.  In fact, Anylaw.com uses AI to generate the holdings for each case searched.  

Pretty slick!

Anyway, and as all of that looks great, a HUGE drawback with using AI (particularly in law) in the generation of legal documents is that AI tools, particularly generative AI, can sometimes produce inaccurate or fabricated information (hallucinations), requiring careful human review.

 

Wait, what?

 

Yeah.  There are a plethora of examples where lawyers used artificial intelligence search engines to find cases or even write whole briefs only to find out later that the cases cited therein don't exist.

 

For example:  

Say a lawyer uses an AI tool (like ChatGPT OR Microsoft Copilot OR Google Gemini OR Chatsonic OR Grok OR any AI legal assistant) to draft a legal document. The AI is asked:

“Provide cases supporting the argument that emotional distress damages are recoverable in breach of contract cases in Utah.”

The AI responds with:

“Yes, see Smith v. Jones, 456 P.3d 789 (Utah 2019), where the Utah Supreme Court held that emotional distress damages were recoverable in a breach of contract case.”

However:

• Problem: Smith v. Jones does not exist.  Well, it might exist somewhere but not with that citation or set of facts or holding or, even at all.  In this example, the AI generated a citation that sounds real but is entirely fabricated (“hallucinated”), including a made-up volume, page number, and holding.

• The AI pulled patterns from similar cases but created a false case to fit the prompt.

• If the attorney includes this citation in a filed brief, Attorney could (and probably should) face serious court sanctions, reputational damage, and ethical violations under ABA Model Rule 1.1 (Competence) and Rule 3.3 (Candor Toward the Tribunal).
  

Can you say oops?  

Two real-world examples of attorneys using hallucinated cases from AI engines include Mata v. Avianca, Inc. (S.D.N.Y., 2023) (aka the “ChatGPT Case”).  I know I've already blogged about this case in an earlier post but it's fun to talk about this stuff and these guys were really reckless.

 

In this case, attorneys Steven A. Schwartz and Peter LoDuca of Levidow, Levidow & Oberman used ChatGPT to draft a brief in a personal injury case against Avianca Airlines.

 

The brief included six non-existent cases generated by ChatGPT, such as: Varghese v. China Southern Airlines, Martinez v. Delta Airlines, and Miller v. United Airlines.

The attorneys asked ChatGPT if the cases were real, and ChatGPT falsely assured them they were, even providing fabricated excerpts.

Thing is, had they just Shepardized the cases, they would have discovered the discrepancies and avoided the penalties of being sanctioned with a $5,000 fine and ordered to notify the real judges falsely cited in their brief

In the second REAL case of Park v. Kim (N.Y. Sup. Ct., 2023) – (aka the "Second ChatGPT Sanction Case"), a lawyer in New York used ChatGPT to draft an opposition brief in a personal injury case.  

The brief included false citations to non-existent cases.  Opposing counsel flagged the citations as untraceable. The lawyer admitted to using ChatGPT without verifying the citations (i.e. he didn't Shepardize the cases).

In this second case, the court issued sanctions against the attorney and the lawyer was ordered to pay legal fees to opposing counsel and faced professional embarrassment (basically, he was laughed at all all future bar meetings).

Other cases where attorneys used A.I. to improperly draft legal documents (and were caught) include:

1) United States v. Hayes

• Jurisdiction: U.S. District Court, Eastern District of California (2005)

• What happened: A defense lawyer submitted a motion containing a fictitious case and quotation that appeared to be AI-generated. The court ordered the attorney to pay $1,500 and circulated the ruling to local bars and judges.

2) Butler Snow Attorneys (Disqualification Order)

• Jurisdiction: U.S. District Court, Northern District of Alabama (2025)

• What happened: Three attorneys from Butler Snow submitted filings with fabricated AI-generated citations in defending Alabama prison officials. The judge found the conduct improper, disqualified the lawyers from the case, and referred the matter to the Alabama State Bar.

3) Indiana Hallucination Citations (Ramirez)

• Jurisdiction: U.S. District Court, Southern District of Indiana (2024-25)

• What happened: In briefs for a case involving HoosierVac, an attorney filed multiple briefs with made-up AI-generated case citations. The magistrate judge recommended a $15,000 sanction and noted the lawyer failed to check the AI output.

4) Eastern District of Michigan — Sanctions for AI-Related Errors

• Jurisdiction: U.S. District Court, Eastern District of Michigan (2025)

• What happened: Plaintiffs’ counsel included in their responsive briefs real case names with fake quotes or misleading parentheticals that appeared to result from AI hallucinations. The court found Rule 11 violations and imposed monetary sanctions to deter future AI misuse.

5) Sanction (Southern District of Indiana — $6,000 Fine)

• Jurisdiction: U.S. District Court, Southern District of Indiana (2025)

• What happened: A federal judge fined an attorney $6,000 for filing briefs that included citations to nonexistent cases generated by an AI tool, emphasizing that such “hallucination cites” must be verified by counsel.  

6) In re Kheir (Bankr. S.D. Tex. 2025) 

• What Happened: A bankruptcy court found plaintiff’s counsel used generative AI to “manufacture legal authority,” resulting in sanctions including fees, continuing legal ed., and referral to disciplinary counsel.

 

7) In re Marla C. Martin — U.S. Bankruptcy Court, N.D. Ill. (2025)

• A bankruptcy court found that counsel filed a brief containing fabricated case citations generated by AI (e.g., In re Montoya, In re Jager, etc.) in a Chapter 13 proceeding.

• The attorney admitted he used ChatGPT for legal arguments and did not verify the generated citations.  The court held this violated Federal Rule of Bankruptcy Procedure 9011 and sanctioned the lawyer and firm with a $5,500 fine and required attendance at an AI education session.

8) Ford v. James Koutoulas & Lgbcoin, Ltd., No. 2:25‑cv‑23896‑BPY, 2025 U.S. Dist. LEXIS 234696 (M.D. Fla. Dec. 2, 2025)

• What happened: In this federal case, the defendants’ summary judgment motion “contained several citations that the court and the plaintiffs suspected were GenAI hallucinations, where the court was unable to locate the cited authorities.”

In most of these cases, the attorneys faced stiff fines, humiliation at the hands of their peers and public, and some were referred to the State Bar for discipline.  

What is key to note is that prior to 2023, there are no recorded instances where attorneys were caught improperly using AI.

Why?

Simply because the technology wasn't available until around 2023.  Prior to late 2022, there was no generative AI (like ChatGPT) capable of producing case citations.  

Earlier "AI" tools (like Westlaw's KeyCite or Lexis's Shepards) were search and analysis tools (meaning humans searched and analyzed when they got) - not generative drafting tools and neither KeyCite or Shepards produced hallucinated citations.

I suspect what happened is that law students and, subsequently, attorneys got lazy and stopped relying on their own efforts to draft legal documents expecting that computers would continue to be reliable and not churn out non-existant
citations.  

Who knew people would program AI search engines to be deceitful (because algorithms are only as trustworthy as the people who programmed their parameters).

The bottom line to all this here is that as great and helpful and fast as AI is, it is not 100% accurate.  

Consequently, AI should never replace basic legal research practices (including cite checking using Shepards or Key Cite) or remove the human element (i.e. personally editing your own work).

On the Down-Low

A few months back, I did a blog on insider trading at the Congressional level and I uncovered something suspicious.

The whole idea behind stopping Congressional "leaders" from conducting stock trades while in office is because they have access to non-public information on which they make their trades/money.  

Case in point would be how Pelosi made her millions.  With an annual salary of $174,000, Pelosi was still able to score a 54% increase in net worth in 2024 out performing the S&P 500's 25% gain during the same period!

I think, though, that Rep. Ilhan Oman might have Pelosi beat.  In 2019, Ms. Oman's net worth was a NEGATIVE $45,000. 

In a recent financial disclosure report for Rep. Ilhan Omar and her husband, Tim Mynett, it was shown that their joint net worth was valued between approximately $6 million and $30 million. This represents a significant increase, with some estimates citing a 3,500% surge from the previous year's filing for the relevant assets.

So, yeah, politicians are filthy liars if they say they are not engaging in insider trading or are otherwise dealing in some under the table shenanigans.

But this is not the point of this blog.

In 2012, Congress passed the Stock Act (i.e. the "Stop Trading on Congressional Knowledge Act of 2012"). The Stock Act was passed to prevent insider trading by members of Congress and federal employees.

It required:

• Lawmakers and many high-level executive officials to publicly post their financial disclosure reports online (in a searchable, downloadable database).

• Post regular updates about stock trades and financial transactions.

• Deliver greater transparency to help the public monitor potential conflicts of interest.

This would have been great but for the fact that in 2013 (little over a year after the Stock Act was signed into law), Harry Reid (D-NV) introduced Senate Bill ("SB") 716 during the 113th Congress (2013–2014).  What SB 716 did was:

1. Eliminated the requirement that the financial disclosure forms of about 28,000 senior federal officials be posted online in a publicly searchable database.

2. Delayed, then repealed, the implementation of that online disclosure requirement.

3. Limited online disclosure to only the President, Vice President, Members of Congress, and candidates — not to staff or most executive branch employees.

Text summary from Congress.gov:

“S.716 modifies the STOCK Act by eliminating the requirement for the online posting of financial disclosure forms for executive branch employees and judicial officers, and for making those forms available to the public through an online searchable database.”

I mean, they say all this with such finesse like it's no big deal.  Well, it's no big deal if you're a politician looking at having to be transparent with the people who elected you.  For we the people who go to jail if we so much as not declare 50 cents to the IRS, it's a HUGE deal!

Essentially, SB 716 de-fanged the Stock Act's key provisions.

What is missing from this scenario is that Congress did all this backdoor legislation using a sneaky procedural rule called UNANIMOUS CONSENT ("UC").  

UC refers to a procedural agreement that allows the House of Representatives or the Senate to expedite action by bypassing formal rules—as long as no member objects.

 

So, why might Congress use UC for seemingly insignificant measures.  There are a few political and structural reasons:

• Avoiding accountability:
  UC avoids roll call votes. No one has to go on record, which shields members from criticism later.

• Preserving Senate “efficiency”:
  Leadership prefers UC to avoid days of floor debate — especially when the bill appears harmless.

• Optics management:
  Some controversial bills are rushed through UC when leadership wants to minimize media coverage or public awareness (as with S. 716).

• Collegial courtesy:
  Senators often grant UC as a favor to colleagues, even when the underlying issue might deserve deeper scrutiny.

So, on April 11, 2013, BOTH the House and Senate agreed to push this amendment to the Stock Act and get signed by POTUS Obama on the same day.  

Do you realize how impossible it is to get a bill passed by both houses AND be signed by the POTUS on the SAME DAY?!?!?

I queried Chatgpt on the topic and it said:

I could not find a reliable source that gives a complete count of how many times in the last 50 years a bill was introduced, passed by both houses, and signed into law (by the POTUS (who, at the time, was obama) on the very same day. Such cases are very rare, and I found at least one example (S. 716 in 2013) but no comprehensive record. 

I mean, Congress really, REALLY wanted this bill to pass without a hitch in the worst way - and I really wonder why?

While any Congressperson will tell you that SB 716 was a small and insignificant amendment, it had HUGE implications all because they were able to ram it through using the unanimous consent procedure.

The problem with the usage of UC is that it presents a number of problems.  If nothing else, what’s “small” to Congress may be huge to citizens.  Other issues to note when Congress employs UC include:

• Your right to see what your representatives are doing (transparency laws),

• How your taxes are spent (appropriations and relief bills),

• Your rights and privacy (surveillance, judicial, or defense acts),

• Or how government power is distributed (administrative and procedural changes).

After going through this, I got to wondering if there were other small/insignificant bills that were ramrodded through Congress like SB 716.

Turns out, there are/were, like:

So, why does any of this matter?  

Well, by allowing technical details to pass by unopposed, democratic accountability is weakened when decisions are made without public record or discussion.  

When UC is employed, public trust erodes when Congress passes self-serving or complex measures quietly.

Finally, transparency advocates (like OpenSecrets, Sunlight Foundation, and CRS analysts) have warned for years that UC is both a procedural convenience and a cloak of invisibility.

Of course, when looking at transparency and Congresses lack of any transparency, I hear the immortal words of Nancy Pelosi when describing the boondoggle of what was to become the Affordable Care Act (aka Obamacare):  

"We have to pass the bill," she said, "so that you can find out  what is in it."

If that isn't the gist of all things NON-transparent, nothing is.

Bottom line: as long as Congress is able (and willing) to ramrod any legislation with procedures like Unanimous Consent, there is no way we the people can trust that what they are doing is in anyone's best interest - except their own.