The Curious Case of AI within the Court of Law
There is a story making its rounds in legal circles that every enterprise leader using AI tools needs to read. Not because it is about lawyers, but because it is about all of us. In short, AI did the work, the lawyer signed the brief and guess who got sanctioned.
In a recent New York appellate case, Deutsche Bank Trust Co. Americas v. LeTennier, attorneys submitted briefs loaded with case law citations. More than 20 of those cases were fictitious, made up entirely. Others were real cases that had been misquoted or mischaracterized, their holdings twisted to support arguments they never actually stood for. The source of the errors was AI tools used to generate the legal research, with no one fact-checking what came back.
The court was direct. Lawyers have a non-delegable duty to verify everything they submit, so using AI is not a defense for false citations. The attorneys were sanctioned, ordered to explain their conduct, and publicly called out in the written opinion. The entire argument was built on fake authority, and the professionals who signed the briefs owned every word of it.
Who is Actually Responsible in this Case?
The reflexive take is to treat this scenario as an AI cautionary tale. I think that misses the point.
AI did not commit the malpractice; the professionals using it did. The tools generated plausible-sounding output, including citations that looked real and read well. Anyone who has spent time with these tools knows this happens. AI hallucinations are not a secret, and it is not a bug that is going to go away. It is a well-known behavior of the technology, and every person using it has the responsibility to understand that.
The failure was not in the use of AI; it was in treating an AI output as a comprehensive final product.
I see versions of this same mistake across industries every week. Not at the scale of fabricated court filings, but in content that goes out unreviewed, analysts that skip a second look, and summaries that get forwarded as fact. The stakes vary, but the pattern does not.
Speed Removed The Friction: The Same Friction That Was Doing Something
When you write something yourself from scratch, the act of writing is also the act of thinking. You catch errors because you are building each sentence deliberately. When AI produces a draft in 30 seconds, you get speed, but what you lose is the friction that was quietly doing quality control.
Errors made in AI outputs do not announce themselves in obvious ways. Instead, they arrive inside fluent, confident, well-structured texts. The faster you move through that text, the easier they are to miss.
The professionals I see doing well with AI have not reduced their review process. They have changed what they are reviewing for. They treat an AI output as a first hypothesis, not a final answer. They bring their domain knowledge to an output before it leaves their hands. That shift in mindset and operational processing is not automatic; it takes deliberate practice.
What A Human Review Actually Requires
"Human in the loop" has become a checkbox phrase. In most organizations, it means someone looked at the output before it went out; that is not a review; it is a glance.
Real reviewing means applying the judgment that only a trained professional can apply. Of course, what that looks like depends a lot on the job.
For a lawyer: verify every citation, not a sample: every single one.
For a marketer: know your brand voice well enough to catch when AI quietly misrepresents it.
For an analyst: understand your data well enough to spot when an AI-generated conclusion is technically possible but contextually incorrect.
For anyone in media: your byline is a warranty. AI does not share that liability you do.
The Courts Are Paying Attention, So Should You
The LeTennier case is not an isolated incident. Since 2023, multiple attorneys across the country have been sanctioned for submitting AI-generated citations that turned out to be false. Courts are now checking citations themselves, and judges are naming AI misuse explicitly in opinions. Some jurisdictions require attorneys to file disclosure statements certifying that citations were independently verified.
The message from the bench is consistent: "I did not know" is not an excuse. "AI produced it" is not a defense. Professional responsibility does not transfer to the tool. Enterprise leaders should be drawing the same line internally, before a client, a regulator, or a court draws it for them.
Quality Is Starting to Matter Again
I have spent 25 years in technology media watching waves of new technology reshape professional work. The pattern is always the same. Early on, the advantage goes to whoever moves fastest. Later, it goes to whoever moved best.
We are somewhere in that transition right now. Speed still matters, but organizations that built real verification habits into their AI workflows from the start are pulling ahead of the ones that treated AI as a volume machine and figured review was someone else's problem.
The attorneys in these cases were trying to move faster. Yes, the AI helped, until it did not. The cost of what seemed like more efficiency was not just a lost filing; it was sanctions, public rebuke, and a tarnished ethics record.
AI does make you faster, but your expertise together with that makes for something worthwhile. Do not outsource the part that requires you.
Posted by Daniel LaBianca on 06/18/2026