In 1966 Stephen Stills wrote about a curfew fight on the Sunset Strip, and the song outgrew its occasion immediately. Sixty years on, nobody remembers the ordinance and few remember the “happening”. What survives is the opening couplet, which names a specific and uncomfortable condition: not ignorance, but knowing you are missing something, with no way to go get it.
That condition now describes the jury box. Jury trial rests on an honor system — twelve people promise to decide on the evidence admitted in the courtroom and nothing else, and we take them at their word. What has changed is that the two most tempting ways to break that promise, asking a machine what it thinks or taking a financial position on the outcome, now sit in the juror’s pocket, cost almost nothing, and leave almost no trace the law permits us to look for.
What It Is Ain’t Exactly Clear
Roughly half of Americans reported using an AI chatbot this year, up from a third two years ago. Somewhere in that number is your next panel. In June, in Paulding County, Georgia, a prospective juror researched a murder case online and shared what he found with the rest of the pool; the judge held him in contempt, released the entire panel, and reset the trial for November. That one is instructive mainly because the juror was caught.
The quieter version is the one that worries me. A juror goes home after a day of expert testimony and asks an assistant to explain postpartum psychosis, or what “reasonable doubt” means as a percentage. A search engine returns links he knows are strangers’ opinions. A chatbot returns a confident, well-organized, lawyerly answer, sometimes citing authority that is from the wrong jurisdiction, superseded, or invented. We tell jurors not to research the case. A juror asking an assistant a general question about the medicine or the law may not think that is research at all — it feels like asking a knowledgeable friend, which is the one thing we forbid, in the one form we never named.
Nobody’s Right If Everybody’s Wrong
Now the money. Prediction markets have gone from novelty to category, and court cases are one of the categories. During deliberations in the recent Lindsay Clancy trial in Massachusetts, Polymarket carried a market on the verdict — roughly $53,000 in volume, conviction trading near thirty-five percent. Whether that is lawful gambling is being litigated right now, and Arizona is in the middle of it: in March the Attorney General filed twenty criminal counts against Kalshi, including operating an illegal gambling business, while the company maintains that federal commodities regulation preempts state gaming law. That fight will take years. The markets are open in the meantime.
The Clancy jury deliberated some forty hours across six and a half days and stood eleven to one for not guilty by reason of insanity. According to the foreperson, the holdout told the room he had reasonable doubt - she says she was so relieved she began filling out the verdict forms - and then added that he still would not return that verdict. Another juror says the man never identified a single piece of evidence or a single witness’s testimony to explain his position. The holdout has now answered them. In a statement released through his lawyer on September 18, he said he had no doubts about Clancy’s guilt and had believed she was guilty from the start, and that the physical evidence and the prosecution’s witnesses convinced him she knew what she was doing and had planned it. He denies ever telling the room he had reasonable doubt, and says that when he tried to work through alternative theories during deliberations he was cut off and treated as though he was expressing doubt.
So we now have two flatly incompatible accounts of the same room. If the foreperson is right, a juror acknowledged reasonable doubt and voted to convict anyway — which is not weighing the evidence differently from his colleagues, but voting against the standard he had just said he applied. If the holdout is right, eleven jurors talked over the twelfth and then mischaracterized him to the press. Both cannot be true, and no lawful mechanism exists that will ever tell us which one is.
The ordinary explanations are almost certainly the right ones on either account — conviction, stubbornness, a room under strain, skepticism of the insanity defense. But an undisclosed financial position would produce the same public record we have: a vote that would not move, and two irreconcilable stories about why. And the economics here are peculiar. A verdict requires persuading eleven other people. A hung jury requires persuading no one. One juror, holding, is the entire trade. I am not accusing this juror of anything. He has now given his own account, and it describes a man persuaded by the evidence; nothing reported suggests that he or anyone close to him wagered a dollar. My point is narrower, and worse: if he had, we would almost certainly never find out.
Everybody Look What’s Going Down
I have watched a private investigator go looking, on the other side of exactly this question. The heading above is an instruction the law declines to follow.
In American Power Products, Inc. v. CSK Auto, Inc., I represented CSK Auto. Twelve days of trial, twenty-four witnesses, 165 exhibits — one of them more than four thousand pages. The jury was out between one and two hours and returned a 6–2 verdict awarding the plaintiff $10,733 on a claim that started at more than $11 million. The plaintiff hired a private investigator, obtained affidavits from two jurors, one of whom stated that a juror had asked the bailiff how long deliberations usually take, and the bailiff, without consulting the judge or counsel, answered that “an hour or two should be plenty.” The plaintiff then moved for a new trial. The trial court denied the motion without an evidentiary hearing; a divided court of appeals reversed. In February 2016 the Arizona Supreme Court reversed the court of appeals and affirmed the denial. (I argued the case.)
The bailiff’s statement was plainly improper. But Arizona Rule of Evidence 606(b) let the plaintiff prove only that it happened (which we stipulated to), not what it did — a juror may not testify about statements made during deliberations, the effect of anything on any juror’s vote, or any juror’s mental processes concerning the verdict. So the court could not ask the eight people who were actually there whether the remark had hurried them. It measured prejudice objectively, against a hypothetical average juror, and found none. Justice Brutinel stated the constraint plainly: a trial court “can never determine with certainty what might or might not have swayed a juror’s decision because direct testimony on that matter is prohibited.”
That is the correct rule. Verdicts have to be final, and jurors have to be able to argue freely without worrying about a private investigator coming to their doorstep afterward. My client won because of that rule, so treat what follows as description and not grievance: the rule cannot tell the difference between a losing party fishing for a second chance and a party with a genuine problem. It turns both away. An improper outside influence is one of the narrow exceptions, and a juror’s wager would qualify — but an exception is only as good as your proof, and the proof cannot come from the twelve people who were in the room. Which is why the chatbot is the more tractable problem. A query sits on a device someone can examine. A wager sits in a market the court has never heard of, in an account that may not carry the juror’s name, and settles quietly a week after the mistrial.
I Think It’s Time We Stop
None of this is cause to lose faith in juries. It is cause to stop relying on an admonition written for a world of newspapers, and to do the work while we still can, because after the verdict Rule 606(b) is waiting. In voir dire, ask the specific question. “Will you agree not to research the case” draws a yes from everyone and tells you nothing. “Do you use ChatGPT or a similar assistant? What do you use it for? If something you heard in this courtroom confused you, would it occur to you to ask it?” gets an honest answer, because most people do not experience that as cheating. Ask the same way about betting apps and prediction markets, and ask about the household, not only the juror. Then ask the court for an admonition that names the behavior, and in a matter with public attention, one on financial interest broad enough to reach a family member’s prediction-market contract.
And raise it when it happens. If a juror’s conduct troubles you — or troubles the other jurors, as it did in Clancy — the moment to make a record is while the jury is still empaneled and the judge can still act. The same concern raised in an affidavit three weeks later is, in most cases, already too late. Clients deserve the translation too: a hung jury costs a full trial’s expense and produces no result, and it belongs in the litigate-or-settle math from the beginning rather than as a surprise at the end.
For What It’s Worth
The song never resolves what is happening. It only insists that something is, and that the sensible response is to stop and look around before the thing plays itself out.
We hand our disputes to twelve strangers, tell them the rules, and trust them to follow the rules when no one is watching. That trust is still well placed, and I have spent close to forty years watching juries mostly deserve it. What has changed is not the character of jurors; it is that the two easiest ways to break faith now cost nothing and feel like nothing — an assistant that will answer anything in the voice of an expert, and a market that will take a position on the verdict for the price of lunch. The rules of evidence, for good reasons, will not let us look afterward. So the looking has to happen first, in the questions we ask and the instructions we request.
This post is part of an ongoing series on how AI is reshaping litigation and the businesses that live through it. An earlier post took up the related question of how to price a dispute when deciding whether to litigate or settle. Earlier installments are collected on the LegaleAZ blog.
A Word About Silver Cain
Silver Cain PLC represents businesses in complex commercial and real estate litigation in Arizona and beyond. When Rebecca Cain and I founded the firm, we built it around direct partner involvement and senior trial-level judgment — including the unglamorous parts of a trial, like voir dire, jury instructions, and the post-trial motions that follow, where cases are quietly won and lost. If the questions in this post are relevant to your business, or to the firms you retain, we are glad to have that conversation.Leon Silver is an AV-rated trial lawyer at Silver Cain PLC, focused on commercial and real property disputes since 1989. He argued American Power Products, Inc. v. CSK Auto, Inc., 239 Ariz. 151, 367 P.3d 55 (2016), before the Arizona Supreme Court. Reach him at lsilver@silvercain.com.
