AI and the future of evidence in family court

Imagine a woman in family court presenting a Nest video of her husband threatening her while their young children cling to her legs. They stand in their penthouse living room, with the Golden Gate Bridge visible through panoramic windows behind them. Each face is unmistakable, as is the husband's angry voice.

The wife seeks emergency orders to protect herself and her children. But her husband denies the incident ever occurred. The video, he tells the judge, is a fabrication created by his wife using artificial intelligence.

Let's assume the husband is lying. He did threaten his wife, with their children present. The recording is genuine and he is simply attempting to create doubt in the judge's mind.

Now, let's assume the husband isn't lying. The evidence is completely fabricated by the wife.

I've practiced family law in Silicon Valley for years, and I've seen plenty of disputes over photographs, text messages and recordings. One spouse claims the evidence is taken out of context. The other claims it's only part of the story.

This is where the rules of evidence aptly step in.

Until recently, family courts largely operated on the assumption that a video, photograph or similar recording was reliable evidence. With some exceptions, of course.

Today, that assumption is completely disrupted.

Anyone with a bare minimum of technical expertise can rely on technology to create realistic photographs, replicate another person's voice or generate a video depicting an event that never occurred. I suspect most of us have little appreciation for just how convincing these creations have become.

For family law attorneys, who represent individuals and families rather than corporations with deep pockets and a plethora of technical experts standing by, the implications are staggering.

How can an ordinary family law litigant afford to prove that a recording is genuine--or that a convincing fake is just that? What happens when a spouse falsely claims that genuine evidence was fabricated? And how much time and money will an innocent party have to spend establishing that their evidence is what it appears to be?

We're about to find out.

The perfect excuse

Almost always, family law litigation comes down to credibility. One party says one thing. The other party says something else. Spouses give completely different accounts about the same incident, as if they're not even describing the same event. Like they weren't even in the same marriage.

As always, the question is: who's telling the truth? In family court, that's up to the judge, the arbiter of truth.

But AI introduces an entirely new possibility: What if the event we're arguing about never happened at all?

A spouse who once would have been forced to explain an incriminating recording now has another option: deny that it's real. Attorneys may now stand before a judge and argue, based solely on their client's claims, that the opposing party manufactured evidence using AI. And sometimes, the attorney will be right.

A particularly difficult problem for family courts

Family law may be especially susceptible to these problems because so much of the evidence originates with the litigants themselves. Here, we're not always dealing with business records maintained by third parties or surveillance footage obtained from a neutral source. We're dealing with spouses recording one another, photographing their children, preserving text messages and documenting incidents they believe may become relevant to a contemplated dissolution.

A parent who records a troubling exchange with a child may send the video to a friend, save it to the cloud or forward it to counsel. By the time the recording reaches court, it may have passed through several devices or applications. Does that make it unreliable? Not necessarily. But it may make it harder to establish precisely where the recording originated and whether it has been altered.

There's also the problem of timing. In emergency custody matters, decisions affecting children's safety and relationships with their parents cannot always wait months for forensic analysis.

These are not theoretical concerns for the people involved. A judge's decision may determine whether someone can remain in the family home or whether a parent can see their children--sometimes, literally, for months.

Getting the evidence right matters enormously.

Can we still trust what we see?

We've always known that people can lie about what happened. We've always known that photographs can be misleading and recordings can be edited. What we haven't had to confront on this scale is the possibility that absolutely anyone can manufacture a remarkably convincing record of an event that never occurred. And perhaps more troubling, that someone who engaged in misconduct can now plausibly claim that the evidence was manufactured.

I don't pretend to have the answers. Family courts are rapidly working through these issues, and the technology will continue to evolve faster than the rules governing its use. But I do think lawyers need to start approaching electronic evidence differently. We should be thinking about original files, how recordings were created, who had access to them and which experts can authenticate the evidence.

We also need to exercise judgment about when an allegation of AI manipulation deserves serious investigation and when it's simply another litigation tactic. These are families, after all, where the stakes are high and budgets are often limited.

Family law attorneys spend much of their careers asking judges to decide which of two people is telling the truth. It's difficult enough when the evidence consists of conflicting memories and competing versions of events. I never expected that we might reach a point where a judge could watch a video of someone committing an act and still be left wondering whether the act ever happened.

Yet here we are.