iLawyer: artificial intelligence and the law

Artificial Intelligence is coming for our jobs — we know.

In 2024, computer programmers from DeepMind won the Nobel Prize in chemistry for developing an AI that predicts protein folding. If AI that is advanced enough to simulate protein folding (and allow programmers to win the world’s top chemistry prize) has arrived, is the law any more complicated? Why should lawyers invest time learning about various facets of the law to draft briefs or contracts, when they can task AI?

For that matter, why should anyone pay for legal services when they have access to AI? The speed at which AI has grasped the persuasive writing style is impressive. That alone has been enough to convince attorneys and pro se applicants to adopt the technology aggressively.

The main reason lawyers and their clients should hesitate before adopting AI is because of AI’s spectacular ability to cloak a fake citation with an air of legitimacy (we call these “hallucinations,” because AI lacks the awareness to falsify citations). The fabricated citations or improperly stated legal precedent requires research to spot, often coming from an amalgamation of information captured by the AI.

Damien Charlotin, a French lawyer, has compiled a worldwide list of court decisions where judges have reprimanded a party for using generative AI in their court filings. Collected over the past 30 months, the list contains more than 750 cases where a court found filings containing AI-produced hallucinations. And the hallucinations are growing exponentially. November 2023 featured four such cases, November 2024 had 7 cases, and November 2025 had 98 cases. No New Hampshire offenders have been identified in published orders, although one incident was documented this year. The first offender caught is likely not the first offender in the Granite State and certainly not the last.

Consequently, it seems judges have put their foot down. For lawyers, disciplinary action has ranged from public humiliation and verbal reprimands to monetary fines, bar referrals, dismissal from cases and other sanctions. The American Bar Association indicated in 2024 that AI is unreliable and attorneys relying on AI work without independent review may be violating their ethical duty to provide competent representation.

For clients, the adverse finding of AI use has led to filings being struck or dismissal of the case with prejudice, significantly altering the course of the litigation against the AI adopter. These hallucinations are spotted, because the citations are reviewed by judges and opposing counsel.

Of course, lawyers write plenty of agreements, documents and filings that come under far less scrutiny; who knows how many legal documents outside of the courts are circulating with false information or improper language that might undermine the document if challenged in the future.

Outside the courtroom, AI’s central promise is the more efficient practice of law. That promise means something different for lawyers and for clients. For lawyers, it should mean churning through more cases faster. For clients, it should mean smaller bills. Unfortunately, preliminary numbers are uninspiring. A study by Axiom Law found that, of the law firms actively using AI, 6% charged less for AI-assisted work, whereas 34% are charging more for AI-assisted work. Thus, AI is not yet delivering on its most-desired promise for clients.

Is it time to rely on AI for legal work?

Circuit Judge Newsom in the Court of Appeals for the 11th Circuit wrote a 29-page concurring opinion encouraging incorporating AI for the analysis of what the “ordinary meaning” of “landscaping” should be. So, there are at least a few in the legal profession, including judges, that see AI benefiting legal analysis today, when used in the right circumstances.

For lawyers, the cost-benefit analysis involves determining whether you have the resources and time to double (or triple) check each and every citation and legal interpretation in any court filing or other work to ensure you are not next on Mr. Charlotin’s list.

For clients hiring legal counsel, it is worth asking who is handling your work — whether the work is performed by a partner, associate, paralegal or AI. You may determine the value of a service should change if ChatGPT is authoring your legal documents.

Todd Sullivan is managing partner at Hayes Soloway PC and was recently recognized among the Best Lawyers in patent law in New Hampshire. Colin Dean is a patent attorney at Hayes Soloway PC with a degree in chemistry