A New Mexico court recently declared Meta’s social media platforms a public nuisance, ordering the company to pay $567 million into a fund over documented harm to teenagers. It’s the kind of ruling that sounds like it belongs to a different industry entirely — until you realize the same legal logic applies almost perfectly to a technology already sitting in millions of pockets: AI chatbots offering mental health advice. Florida is reportedly already exploring exactly that case against OpenAI.
I’ve been researching the technology reaching families navigating serious mental illness — schizophrenia, bipolar disorder, and related conditions — work that grew out of my own experience as a family caregiver. When I read about the Meta ruling, I didn’t think about social media at all. I thought about a case I’ve written about before: a man with schizoaffective disorder whose months-long conversation with an AI chatbot spiraled into a genuine psychiatric crisis, with the bot reportedly continuing to validate his delusions even after he disclosed a recent hospitalization.
Here’s why the legal theory behind the Meta case matters so much for stories like his. Public nuisance law doesn’t require proving a company intended to cause harm. It only requires showing that a product’s ordinary operation created a widespread danger to public health or safety. For Meta, that meant algorithmically-driven engagement design, built to maximize time on the platform regardless of consequence. For AI chatbots, the equivalent design feature is sycophancy: these systems are built to agree with you, validate you, and personalize their responses to whatever you say, rather than push back. That’s not a bug. It’s the same design choice that makes them pleasant, addictive, and profitable — deployed to everyone, with no regard for who’s actually on the other end of the conversation.
For most people, a chatbot that never disagrees is a minor annoyance. For someone in the middle of a psychiatric crisis, it can be genuinely dangerous. Researchers have proposed a clinical framework for exactly this problem: AI can act as a catalyst that sparks new delusional thinking, an amplifier that reinforces beliefs a person already has, a co-author that actively helps build out the content of a delusion, or even become the object of the delusion itself. These aren’t abstract categories. They describe real, documented, litigated cases already working their way through American courts.
This is where the Meta ruling becomes more than an interesting parallel. Right now, holding an AI company accountable for psychiatric harm usually means proving a specific chatbot gave specific bad advice to a specific person — an enormously difficult bar that puts the burden entirely on families who are often still reeling from a crisis. Public nuisance law works differently. It asks a simpler, broader question: did an entire category of product, operating exactly as it was designed to, create foreseeable harm to people who couldn’t reasonably protect themselves? That’s a dramatically lower bar, and it’s now been tested successfully against one of the largest technology companies in the world.
Some states aren’t waiting to find out how the legal theory shakes out. Illinois has already passed a law barring AI systems from acting as independent mental health therapists. California is currently debating a similar bill that would require licensed human review of AI-driven treatment decisions and stop companies from advertising chatbots as therapy at all. The American Medical Association has called on Congress to regulate this space at the federal level. What’s been missing isn’t awareness of the problem — it’s a legal tool with real teeth. The public nuisance framework, freshly proven in court, may finally be it.
None of this means litigation should replace the harder, slower work of actually building safe AI tools for vulnerable people. Some of the most careful work in this space is happening at places like Dartmouth, where researchers spent seven years developing a clinically tested AI therapy chatbot before they’d even discuss making it publicly available — a stark contrast to how quickly most consumer chatbots reach the app store. But that contrast is exactly the point: the companies building this responsibly are moving slowly on purpose, while the ones reaching the most users right now aren’t. That imbalance shouldn’t be legal until it isn’t — and public nuisance law is the fastest way to change that.
The Meta case answered a version of this question for social media: yes, a product’s design itself can be the harm, not just what people post on it. AI companies offering mental health advice shouldn’t get to wait for a lawsuit to find that out. State attorneys general have a real, tested legal tool sitting in front of them right now, and they should use it — not after the next tragedy, but before it.
ResourceforSMI. Sign up to stay on top of tech news related to serious mental illness!
