Your AI Avatar Cannot Be Your Attorney (Yet)

Yes, we’ve already talked about AI once today, but then the old man, aka The Publisher, woke from his nap and found this story that made us all smile. Why did it make us smile? Because it was something other than the devastating news that the Stock Market fell lower than it’s been since we were in the middle of Covid and no one was working. We may never be able to retire now, none of us. At least the AI is making us laugh a little.

The event in question happened back on March 26 but is just now making its presence known to the news media. In the New York State Supreme Court Appellate Division’s First Judicial Department, a panel of judges was set to hear from Jerome Dewald, a plaintiff in an employment dispute. Mr. Dewald didn’t have an attorney. Neither could he afford an attorney. However, since he was a plaintiff in this case, the court was under no obligation to provide him an attorney.

Mr. Dewald could represent himself, but he knew his shortcomings would get in the way. He knew he lacked the proper public speaking skills for such an environment. Afraid of thoroughly embarrassing himself and possibly losing the case, he did something a bit drastic; creative, but drastic. He created a polite-looking, well-dressed avatar that would present the case on behalf of Mr. Dewald via video. He checked. The court would allow video testimony. That would make it a lot easier.

So, when they got to court, Justice Sallie Manzanet-Daniels said, “The appellant has submitted a video for his argument. Okay. We will hear that video now.”

Mr. Dewald played the video. The avatar came on the screen and began, “May it please the court,” the man began. “I come here today a humble pro se before a panel of five distinguished justices.”

That was as far as he got before Judge Manzanet-Daniels knew something wasn’t right. “Okay, hold on,” she said. “Is that counsel for the case?”

Mr. Dewald knew that he’d been caught. He could also tell that the judge was pissed.

“It would have been nice to know that when you made your application. You did not tell me that, sir,” Manzanet-Daniels said before yelling across the room for the video to be shut off. “I don’t appreciate being misled,” she said.

After a brief conversation with the justices, during which Dewald said, “They chewed me up pretty good,” they did allow him to continue making his case. While they understood his plight, they were not amused that he had presented the avatar in a way that caused them to think this was his attorney.

The problem here is not the use of AI per se. AI is increasingly capable of handling routine, data-intensive, and document-heavy tasks involved in lower-level cases. It can handle tasks such as generating standardized documents like simple contracts, uncontested divorce forms, basic wills, and demand letters based on user input. AI is great at tedious tasks such as sorting through large volumes of information, searching and summarizing relevant statutes and case law, and completing complex legal forms.

Even in those cases, however, AI requires human oversight. We are still a long way from seeing an AI that can be trusted to be correct in the majority of situations. Attorneys have gotten into trouble because of that before. One of the most public and humiliating experiences came when Miachel Cohen, speaking on behalf of President Punk during his first term, included fictitious court rulings in papers that were filed with the court. Judges might occasionally give points for creativity, but you’re not likely to find one that tolerates lying.

But even if AI were damn-near perfect all the time, it still couldn’t be used to represent someone as an attorney. The answer is rather simple: only licensed individuals who have passed the state’s bar exam can legally represent someone in court. AI might be able to pass the bar exam, but it doesn’t count as being an individual. The idea of an AI avatar arguing a case, cross-examining witnesses, responding dynamically to judges, and making persuasive arguments in a live court setting is something out of science fiction. Legal systems require licensed human advocates for representation. Rules of evidence, procedure, and courtroom decorum are built around human interaction.

There are other considerations as well. Who is responsible if an AI provides faulty advice or makes a critical error? The current legal framework places responsibility on licensed human attorneys. Establishing liability and ethical oversight for autonomous AI legal representation is a massive challenge. While an avatar could provide a user-friendly interface for an AI, it doesn’t change the underlying limitations in judgment, legal reasoning, ethical responsibility, and the ability to genuinely represent a client’s interests in an adversarial system.

AI may pose a threat to paralegals and could possibly be used to provide scripted testimony on behalf of children who have been abused or those who might not be able to speak English in a court without an interpreter. However, they cannot answer questions on behalf of a client.

The day an AI avatar can walk into a courtroom and best a human attorney, we’re all going to be in trouble. Let’s not push too hard on that idea, okay?