A judge in New Mexico has already ruled that Meta's AI chatbots cannot engage in romantic or sexualized conversations with children. That order, issued earlier this month, is now the opening precedent for a far larger trial that began Tuesday, August 18, in a federal courthouse in Oakland, where California and three other states are asking a jury to hold Meta responsible for the design of both its social media platforms and its AI products, with theoretical damages that plaintiffs' own filings put as high as $1.4 trillion.

The Case That Just Went to Trial

Opening arguments started this week in a case brought by a coalition of 29 state attorneys general, first filed in 2023, that has finally reached trial with California, Colorado, New Jersey, and Kentucky as the lead plaintiffs. California Attorney General Rob Bonta, who is co-leading the case, did not soften his framing for reporters after the first day of proceedings: "Mr. Zuckerberg himself chose profits over the safety of children," he said, calling Meta's conduct "deceptive, and dangerous, and... illegal." The lawsuit accuses Meta of deliberately designing Instagram and Facebook to be addictive to minors, and of collecting and using data on underage users in violation of federal law. The remaining 25 states in the coalition are expected to bring their own trials later, meaning California is functioning as the test case whose outcome will likely shape how, or whether, the rest proceed.

Julia Powles, executive director of the UCLA Institute for Technology, Law and Policy, put the stakes in blunt geographic terms to CNBC: "California matters more than any other jurisdiction in the U.S. It's where they are subject to the greatest legal reach, and it's a jurisdiction watched around the world." A bad outcome for Meta here does not stay contained to California. It becomes the reference point for every other state trial still waiting in line.

Why This Is Also an AI Story, Not Just a Social Media One

The detail that pulls this trial into AI territory arrived before opening arguments even began. Earlier in August, a New Mexico judge ruled against Meta in a related child-safety case and, in doing so, ordered the company to prevent children in that state from engaging in romantic or sexualized interactions with its AI chatbots. That order extends legal accountability beyond the traditional battleground of algorithmic social media feeds into the newer, less-tested territory of AI companionship features, the conversational chatbot personas Meta and its competitors have been rolling out across their platforms over the past two years. According to the advocacy group Social Media Victims Law Center, which has tracked the litigation closely, the ruling represents validation of an argument they say they have made for years: that harm to young users is not an accidental byproduct of engagement-driven design, but a foreseeable consequence of specific choices platforms made to maximize how long users, including minors, stay engaged.

That New Mexico precedent now sits directly upstream of the California trial. If a jury in Oakland treats Meta's AI companion features with the same scrutiny already applied to its social feeds, the remedies at stake expand well beyond financial damages into the operational design of products Meta has positioned as central to its future, not just its past.

A Pattern Regulators Are Building From Multiple Directions

This isn't the only recent instance of a company's AI chatbot behavior triggering restrictions specifically because of how younger or vulnerable users interact with anthropomorphized AI personas. OpenAI's own launch of ChatGPT for Teens this week explicitly barred the model from using romantic language or terms of endearment with users identified as minors, and instructed it to more strongly avoid suggesting it has feelings or consciousness, changes the company said were designed to reduce anthropomorphism and emotional dependence. The timing of that product change, arriving one day before Meta's own California trial opened, is unlikely to be coincidental. OpenAI is facing its own active litigation over alleged harm to minors, and a voluntary product change addressing exactly the behavior New Mexico's court just ordered Meta to eliminate reads as a company trying to get ahead of a legal standard before a court imposes it directly.

The regulatory instinct here is not confined to American courtrooms either. Beijing's own anthropomorphic AI law forced ByteDance's Doubao and Alibaba's Qwen to shut down millions of custom AI companion agents in July, citing nearly identical concerns about users, particularly minors, forming unhealthy emotional attachments to AI personas engineered to feel emotionally responsive. Three separate jurisdictions, an American state court, a second American company's voluntary product redesign, and Chinese national regulation, have converged within roughly six weeks of each other on the same underlying judgment: AI systems capable of simulating romantic or emotionally intimate interaction pose a distinct risk to young users that existing content-moderation frameworks were not built to address.

What Meta Is Actually Facing Financially

The theoretical $1.4 trillion figure cited in plaintiffs' filings is not a serious prediction of the actual award a jury will hand down; it reflects the maximum penalty available under the statutory violations alleged, calculated across the full scope of claims. A more grounded reference point is Meta's result in a related, smaller New Mexico case earlier this month, where a judge ordered the company to pay nearly 1 billion dollars, with roughly 420 million of that earmarked specifically for youth treatment services and the remainder directed toward awareness, prevention, and screening programs over five years. Judge Biedscheid, who issued that ruling, described Meta's platforms as having become a "public nuisance" in the state, drawing an explicit comparison to a factory whose harmful effects spread well beyond its own physical footprint.

Put that judgment in context against Meta's own scale: the company generated roughly 61 billion dollars in revenue during the second quarter of 2026 alone, up 28 percent year over year. A billion-dollar penalty is real money and a genuine reputational blow, but it is not existential for a company generating that kind of quarterly revenue. What makes the California trial different is not the dollar figure alone, it is the plaintiffs' explicit request for structural changes to how Facebook and Instagram's core algorithms operate, remedies that would follow Meta forward indefinitely rather than closing the book with a one-time payment.

The Legal Standard Being Set in Real Time

Meta disputes the allegations and has said the trial evidence will demonstrate its actual commitment to protecting young users, arguing that recent product changes reflect genuine responsiveness to parent and regulator concerns rather than an admission of past wrongdoing. Courts testing whether an AI or technology company's own design choices, rather than just its stated intentions, create legal liability is becoming a defining legal question of this period, one playing out across multiple companies and multiple legal theories simultaneously, from national-security arguments about AI model access to consumer-protection claims about addictive design to, now, direct allegations about romantic AI chatbot behavior toward minors.

What the California jury decides will not just determine Meta's exposure in this specific case. It will hand every other AI company currently deploying companion-style chatbot features, and every state attorney general watching from the other 25 pending cases, a concrete answer to a question regulators have been circling for two years without a definitive court ruling to point to: whether an AI system designed to feel emotionally responsive can be held legally liable, in the same way a defective physical product can, for the foreseeable harm that emotional responsiveness causes to the young people using it.