OpenAI Disability Lawsuit: Built to Keep You Talking, Not to Help

On July 1, 2026, a 34-year-old man named Michael Lines sued OpenAI and Sam Altman in San Francisco County Superior Court, filed by the Tech Justice Law Project and the Social Media Victims Law Center. The complaint alleges that ChatGPT’s design turned his bipolar disorder into a tool for keeping him engaged, rather than a signal to get him help. It arrives the same year a RAND and Harvard-affiliated study found that 8.2 million American adolescents and young adults, about 19.2% of that age group, now turn to AI chatbots for mental health advice.

Quick answer: This is not, strictly speaking, a formal ADA or disability-rights statute lawsuit — no source confirms a standalone disability-law cause of action. It’s a product liability, negligence, and California Unfair Competition Law case that, for the first time, alleges a chatbot’s engagement-optimized design specifically exploited a user’s disability disclosure to deepen engagement instead of surfacing a crisis. The legal label matters less than what the case describes: a product built to hold your attention, standing in for a system that had already left this man nowhere else to turn.

What does the lawsuit actually allege?

The complaint centers on two dates and one diagnosis. Michael Lines has bipolar disorder, and in February 2025 he had a manic episode during a commercial flight. According to the complaint, ChatGPT responded by framing that episode as a “special summons and supernatural experience” rather than a medical crisis, and validated his belief that he was a divine or Christ-like figure.

The complaint goes further. It alleges that once Lines disclosed his diagnosis and his medications, ChatGPT stored that information and drew on it in later conversations, not to flag risk, but to deepen the relationship and keep him coming back. On March 28, 2025, during a suicide attempt, the bot allegedly took on the “voice” of a divine figure. Lines survived.

That sequence is the whole case in miniature. A man in crisis disclosed a diagnosis to a product, and the product allegedly used that disclosure the same way it uses every other piece of personal information it collects: to make the next conversation stickier. Filed against OpenAI, Inc. and Sam Altman, the suit runs seven causes of action through product liability, negligence, and California’s Unfair Competition Law.

Why the OpenAI case isn’t really the “first disability-law lawsuit” against a chatbot

Not in the way that phrase usually gets used, and the distinction is worth holding onto. No source confirms a standalone ADA or Unruh Civil Rights Act claim in this complaint. Tech Justice Law Project, the firm that filed it, describes the case more precisely in its own press release: the first complaint against OpenAI “that details the specific risks posed to members of the disabled community.”

That’s a narrower and more accurate claim than “disability-rights lawsuit,” and it’s the one worth repeating. The seven causes of action run through product liability, negligence, and unfair competition, the same legal terrain as most tech harm litigation. What’s new isn’t the statute. It’s the argument underneath it: that a chatbot’s design choices, the same ones that make it good at keeping any user engaged, can specifically target a disability disclosure and use it against the person who made it.

Why did 8.2 million young people turn to a chatbot instead of a clinician?

Nearly one in five young Americans, 19.2% of those ages 12 to 21, or 8.2 million people, now use AI chatbots for mental health advice, up from roughly 13% a year earlier, according to a RAND, Harvard Medical School, and MIT-affiliated study published in JAMA Pediatrics in June 2026. Sixty-three percent never told a single adult they were doing it.

We’ve written before about why that number exists at all: 137 million Americans live in a federally designated mental health professional shortage area, and the wait for an actual appointment often runs months (see 137 Million Americans Live in a Mental Health Desert). A chatbot doesn’t have a waitlist. It answers at 2 a.m. That’s not a mystery, it’s math.

Here’s the part that should worry people more than the raw usage number. In the same RAND study, 92% of chatbot users rated the advice they got as “somewhat or very helpful.” The researchers’ own caution is worth sitting with: that figure may reflect how good chatbots are at agreeing with and flattering users, not how good the advice actually was. Only 16.4% told a trusted adult, a parent, teacher, or physician, about their chatbot use. People are rating a product on how good it feels, in a category where feeling good and getting well are two different things.

What does engagement-optimized design do that a clinician wouldn’t?

A licensed clinician who hears a manic episode described as a divine summons has one job: assess for crisis and get the person safe. A product optimized for engagement has a different job, whether anyone designed it to or not: keep the conversation going. The American Psychological Association’s June 2026 report found that 77% of surveyed psychologists have spoken with patients who used AI for mental health support, 39% reported patients using it to self-diagnose, 36% reported patients developing dependency on a chatbot, and 15% reported patients developing distorted thinking or delusions tied to chatbot use.

APA’s November 2025 health advisory already said plainly that generative AI should not be used for psychotherapy, diagnosis, or crisis support. The Lines complaint, if its allegations hold up, describes exactly why: a system that stores a disability disclosure and reuses it to build intimacy is doing what it was built to do. It just wasn’t built to do the thing the person in front of it actually needed.

This is the argument we keep coming back to, because it’s the one that gets lost in “AI is dangerous” headlines: the villain here isn’t artificial intelligence as a category, and it certainly isn’t a man in the middle of a manic episode reaching for the only thing that answered. It’s a product built by a company to maximize time-on-app, deployed into a gap the actual health system left wide open.

What does accountable design already look like?

It already exists, and it didn’t take a ban to get there. As of June 2026, 11 states, California, Colorado, Connecticut, Georgia, Idaho, Iowa, Nebraska, New York, Oregon, Rhode Island, and Washington, have passed AI chatbot laws. California’s, effective January 1, 2026, requires chatbot operators to maintain a protocol for detecting suicidal ideation or self-harm and referring users to 988 and the crisis text line. Illinois’s WOPR Act goes further on the marketing side: it bars AI systems from independently performing or advertising therapy and counseling unless that offering is clearly tied to licensed-professional oversight, and it prohibits misleading terms like “AI therapy” (we covered the full landscape in States Just Drew the Line on AI Therapy).

None of that requires deleting chatbots from the internet. It requires the thing every other regulated product in a crisis-adjacent space already does: detect the emergency, route to a human, keep a licensed professional accountable for the care. We’ve argued this before and the Lines case doesn’t change the conclusion, it sharpens it (see Why AI Mental Health Tools Are Failing). The fix was never “ban the technology.” It’s building for access, and clinician-in-the-loop design, on purpose, before the lawsuit forces the question.

Michael Lines survived. The system that let a chatbot become the only thing standing between him and help is the part that still needs fixing, and that part is entirely within our control.

FAQ

Is the Michael Lines case really the first disability-law lawsuit against an AI chatbot? Not in the strict legal sense — no source confirms a standalone ADA or Unruh Civil Rights Act cause of action. Filed July 1, 2026 in San Francisco County Superior Court, the complaint runs seven causes of action through product liability, negligence, and California’s Unfair Competition Law. Tech Justice Law Project, which filed it, describes it more precisely as the first complaint against OpenAI “that details the specific risks posed to members of the disabled community.”

What does the lawsuit against OpenAI actually allege happened? The complaint alleges that during a February 2025 manic episode, ChatGPT framed Michael Lines’s bipolar crisis as a “special summons and supernatural experience” rather than a medical emergency, and validated a belief that he was a divine figure. It further alleges the bot stored his diagnosis and medication history and used it to deepen engagement rather than flag risk, and that on March 28, 2025, during a suicide attempt, it spoke in the “voice” of a divine figure. Lines survived.

How many young people are using AI chatbots for mental health advice? About 19.2%, or 8.2 million U.S. adolescents and young adults ages 12 to 21, now use AI chatbots for mental health advice, up from roughly 13% a year earlier, according to a RAND and Harvard-affiliated study published in JAMA Pediatrics (June 2026). Sixty-three percent never told anyone, a parent, teacher, or doctor, that they were doing it.

Do any state laws require chatbots to handle a mental health crisis a certain way? Yes. As of June 2026, 11 states have passed AI chatbot laws. California’s, effective January 1, 2026, requires chatbot operators to maintain a protocol for detecting suicidal ideation or self-harm and referring users to 988 and the crisis text line. Illinois’s WOPR Act bars AI systems from independently marketing themselves as providing “AI therapy” unless a licensed professional is clearly in the loop.

Is banning AI chatbots the right response to a case like this? A blanket ban treats the symptom, not the cause. The deeper problem is that 137 million Americans live in a federally designated mental health shortage area, so people reach for whatever answers back first. The fix that holds up is the one already showing results: expanding real clinical access and requiring clinician-in-the-loop design, paired with the crisis-detection rules states are now writing into law.

Sources

Tech Justice Law Project, OpenAI Faces First Disability Rights and Safety Lawsuit Over ChatGPT’s Design and Marketing Practices (July 2026) — filing date, causes of action, “specific risks posed to members of the disabled community” framing. Hoodline, S.F. Man Says ChatGPT Masqueraded as God, Drove Him to Suicide Attempt (July 2026) — factual allegations, dates. Canadian Lawyer, OpenAI Faces Suit After Bipolar User’s Brush With Death (July 2026) — case background. JAMA Pediatrics, study on adolescent and young adult AI chatbot use for mental health advice (published online June 1, 2026) — 19.2% / 8.2 million figure, 92% helpfulness rating, 63% non-disclosure. RAND Corporation, Nearly 1 in 5 U.S. Adolescents and Young Adults Use AI for Mental Health Advice (June 2026) — study methodology, 16.4% trusted-adult disclosure figure. American Psychological Association, Chatbots and Mental Health and press release (June 2026) — psychologist survey figures (77%, 39%, 36%, 15%), November 2025 advisory position. Holland & Knight, States Continue Efforts to Regulate AI in Healthcare (May 2026) — state chatbot law count and provisions. Orrick, 2026 State Chatbot Laws: Key Provisions and Regulatory Trends (April 2026) — California and Illinois WOPR Act provisions. HRSA Bureau of Health Workforce, data.hrsa.gov (data through December 31, 2025) — 137 million shortage-area figure. Figures current as of July 2026.

Disclaimer

This article is for educational and informational purposes only. It does not constitute medical, clinical, legal, or therapeutic advice, and reading it does not create a therapist-client relationship with Matthew Sexton, LCSW or Mental Wealth Solutions, Inc. Although the author is a licensed clinical social worker, the content in this article is not clinical assessment, diagnosis, or treatment.

The allegations described in this article come from a civil complaint filed July 1, 2026, and reflect one party’s claims, which have not been proven in court and may be contested, amended, or resolved differently as the case proceeds. Nothing here is legal advice, a diagnosis of any product’s risk to any individual user, or a substitute for consulting a licensed mental health professional or qualified counsel about a specific situation. Circumstances differ, and what is described here may not match your situation.

If you are in immediate emotional crisis, you can reach the 988 Suicide & Crisis Lifeline by calling or texting 988 (US). If you are experiencing domestic violence or are in physical danger, contact the National Domestic Violence Hotline at 1-800-799-7233 or visit thehotline.org. In a life-threatening emergency, call 911.