Last week, prosecutors in Clark County, Nevada, moved for sanctions against a Nevada lawyer,
alleging he filed a brief written largely by artificial intelligence. Apparently the syntax, the
formatting, and a cited case that does not exist were the flags that tipped them off. The lawyer
says he has never used AI and would not know how. Whatever the truth turns out to be, the
filing is now the story.
For those of us in the law, it is a familiar story. In September 2025, a Washoe County judge
sanctioned two lawyers whose brief contained fourteen citations fabricated by ChatGPT. One of
them did not know AI had been used at all — but he had signed the pleading before it was filed.
In the first three months of 2026, a single federal judge in the District of Nevada issued four
separate orders addressing AI-generated citations. We lawyers have spent two years learning
this lesson in public, one sanctions order at a time.
Here is what scares me: while lawyers have been busy embarrassing themselves with AI, the
construction industry has been quietly — maybe not so quietly — putting it into practice at a
level the legal profession has not begun to reckon with. Those uses create transactions that
could, and in some cases will, result in litigation. Litigation runs on rules. And the rules
surrounding AI in evidence and court decisions, right now, largely do not exist.
The on-site implementation
The Wall Street Journal reported this spring that Weyerhaeuser, the largest logger in the
country, is building a digital twin of timberland roughly the size of Indiana — satellite imagery,
drone photography, and lidar mapping the size and species of individual trees. An algorithm the
company built decides which trees come down and which are left standing. A driverless skidder
has already dragged timber around a Southern logging site while its operator sat in a home
office four hundred miles away.
Engineering News-Record reported in June that John Deere, through its Wirtgen Group,
demonstrated a connected roadbuilding operation in which milling, paving, and compaction
machines share production data across a single platform. Quality information flows off the
machines and into a permanent digital record. Company officials were candid about the ambition:
the technology is intended to produce a durable digital record of pavement quality.
They were equally candid that whether transportation agencies will accept digital coring in
place of physical pavement sampling remains an open question.
And in estimating — a task that seems to set up a lot of construction disputes — two
consultants wrote this spring that AI advancement is outpacing industry regulation and creating
unclear expectations about how long proper estimating should take, warning that clients treat
preliminary projections as final answers. Obviously, that was a business problem. But it is also a
source of potential delay claims, waiting in the wings.
The questions no court has answered
Take those three developments, assume a dispute arises from them, and put them in a real
courtroom — on a defect or a delay claim of the kind my firm handles every month.
When an algorithm decides sequencing, materials, or methods on a project, and the project
runs late, what is admissible into evidence? The output? The model? The training data? Who
even controls that data well enough to establish sufficient custody as a business record? In
March of this year, a federal court in Colorado ordered a party to disclose which AI tool it had
used, while protecting the substance of the analysis. That was a document-review dispute.
Apply the same logic to an AI-based schedule optimization tool that hallucinates a bunch of
logic, and you have an evidentiary fight nobody has had yet — at least that we have seen
reported.
When a digital record replaces a core sample, who authenticates it? A core sample is a physical
thing a human being pulled from the ground and can testify about. A digitally generated quality
record is an inference produced by an algorithm, and Deere has already acknowledged that
agency acceptance is unresolved. If the agency has not decided whether to accept it, a court
certainly has not. Is this a new area of consultant expert testimony?
When a contractor feeds project data into an AI tool to analyze a delay claim in anticipation of
litigation, is that work product? The honest answer today is that it depends on which
courthouse you are standing in. In February, a federal court in New York found, in a criminal
case, that a defendant who used a consumer AI tool on his own initiative to draft his defense
strategy had waived privilege — calling it a question of first impression nationwide. Within
weeks, civil courts in Michigan, Colorado, and Texas reached materially different conclusions,
treating the AI as an instrument rather than a living third party, and protecting the work
product. The dividing line so far appears to be criminal versus civil — not who typed the prompt.
No appellate court has resolved the split. In Nevada, there is no state ethics opinion
and no statewide court rule on any of it.
The question that should worry the industry most
In May, a New York court found that a party which had relied on generative AI to screen
roughly nineteen thousand documents for privilege — with no sampling, no iterative
refinement, and no human re-review — had waived privilege over nearly all of them. The
court’s reasoning was not that AI use is improper. It was that unvalidated reliance on AI is a
failure of reasonable care.
Read that sentence again with a contractor in mind rather than a lawyer.
The American Bar Association’s Formal Opinion 512, issued in July 2024, told lawyers they must
understand the capabilities and limitations of any AI tool they use, must independently verify
its output, and remain responsible for the work of anyone they supervise who uses it. That
guidance now has teeth in Nevada, as Washoe County demonstrated.
The construction industry has no Formal Opinion 512. There is no equivalent guidance for an
estimator who accepts an AI-generated number, an engineer who stamps an AI-assisted
drawing, or a general contractor who passes along a subcontractor’s AI-produced submittal.
The absence of guidance is not the same as the absence of a duty. Courts have shown
themselves entirely willing to ask whether a professional validated the machine’s work before
relying on it. There is no principled reason that question stops at the courthouse door and
never reaches the jobsite.
What I think is coming
I do not think AI is a threat to construction, or even to industry jobs. I think it is going to make
projects faster, cheaper, and better documented than anything we have seen, and the firms
that adopt it well will win work from the firms that do not.
But every one of those efficiencies is also a new evidentiary artifact. The digital twin is a record.
The algorithmic cut sheet is a record. The connected paving data is a record. The AI-assisted
estimate is a record. When the dispute comes — and in construction, the dispute always comes
— those records will be demanded in discovery, contested for authenticity, and argued over by
people who have not yet agreed on what any of it means. At the end of the day, a court has to
decide whether it falls into a hearsay exception already on the books, or make up a ruling in
real time. A ruling that could be overturned on appeal.
The contracts being signed on Nevada projects today are largely silent on all of it. They do not
allocate ownership of project data generated by an equipment vendor’s platform. They do not
address whether AI-assisted analysis is protected. They do not say who bears the risk when an
algorithm’s recommendation turns out to be wrong. That silence is not neutral. It is a default,
and defaults get resolved by courts applying rules that were not written with any of this in
mind.
I would rather see these questions answered at the negotiating table, while they are still
contract terms. Because the alternative is answering them in front of a judge who is seeing
them for the first time, in a case where my client’s money is on the line.
—
Leon F. Mead II is the founding attorney of Mead Law Group LLP in Las Vegas and the author of Nevada Construction Law. He has been ranked Band 1 for Construction — Nevada by Chambers USA continuously since 2006.