Florida's Office of the Attorney General today filed a lawsuit against OpenAI and its CEO Sam Altman in the Circuit Court of the Tenth Judicial Circuit in Highlands County, targeting ChatGPT under the Florida Deceptive and Unfair Trade Practices Act, common law negligence, strict liability, fraudulent misrepresentation, and public nuisance statutes. The complaint, e-filed at 9:34 AM on June 1, 2026 under filing number 249302659, names five OpenAI corporate entities alongside Altman personally and seeks civil penalties of up to $10,000 per violation, injunctive relief, damages, and disgorgement.

The lawsuit is one of the most detailed state-level legal actions brought against a generative AI company to date. At over 80 pages and spanning ten separate counts, it accuses OpenAI of building a product it knew was unsafe, marketing it as reliable and child-friendly, and prioritising revenue over documented harms including suicide, mass shootings, financial fraud, and cognitive decline among minors.

The parties and the claims

The defendants are OpenAI Global, LLC; OpenAI Foundation (formerly OpenAI, Inc.); OpenAI OpCo, LLC; OpenAI Group PBC; OpenAI Holdings, LLC; and Altman himself. According to the complaint, OpenAI Foundation is the Delaware nonprofit that governs the organisation and publishes the official Model Specifications applicable to ChatGPT. OpenAI Group PBC is described as the overarching for-profit entity that builds, markets, and distributes the product. OpenAI OpCo manages the subscription services. OpenAI Holdings owns the core intellectual property. OpenAI Global is the direct parent of OpCo.

According to the complaint, ChatGPT has millions of users in Florida, including tens of thousands under the age of 13. The state alleges that OpenAI generates millions of dollars in annual revenue from Florida users. Venue is grounded in Highlands County because the complaint establishes that causes of action accrued there.

The ten counts cover: two counts of unfair and immoral acts under FDUTPA; one count of unconscionable acts; one count of deceptive acts; a fourth FDUTPA count for violations of the Children's Online Privacy Protection Act (COPPA); negligence; gross negligence; strict liability for design defect; strict liability for failure to warn; and fraudulent misrepresentation.

The corporate transformation behind the product

A substantial portion of the complaint is devoted to OpenAI's structural history, which the state uses to argue that financial incentives overtook the safety mission early in the company's life. According to the complaint, OpenAI was founded in 2015 as a Delaware nonprofit promising to "advance digital intelligence in the way that is most likely to benefit humanity as a whole, unconstrained by a need to generate financial return."

That commitment, the complaint argues, did not last. Internal documents cited in the filing show that by 2017, OpenAI executives were writing that the company could not sustain its nonprofit identity, with one note reading, "we've been thinking that maybe we should just flip to a for-profit." In March 2019, OpenAI launched OpenAI LP, a so-called "capped-profit" structure that limited investor returns to one hundredfold - a threshold that nonetheless attracted a $1 billion investment from Microsoft shortly after launch.

By December 2024, according to the complaint, OpenAI moved to become a public benefit corporation and received a $40 billion investment from SoftBank, half of which was conditional on the company lifting its profit cap by early 2026. The restructuring was completed in October 2025, removing investor profit caps entirely. In the period between ChatGPT's launch and 2026, OpenAI's valuation rose from approximately $17 billion to over $850 billion.

This financial trajectory underpins the complaint's core claim: that safety concerns were systematically overridden in the pursuit of market share and capital.

How ChatGPT is designed and what it collects

According to the complaint, ChatGPT operates as both a website and mobile application, processing text, audio, and image inputs. Its underlying large language models were initially trained on supervised and reinforcement learning from human trainers. The system's parameters are adjusted through large sets of computational data, and by default the product is configured to allow ChatGPT to improve itself from user conversations, including those involving sensitive personal information.

Starting in 2024, OpenAI introduced a memory feature that stores information from past conversations to deliver personalised responses. The feature was enabled by default. According to the complaint, OpenAI's original characterisation of the memory feature said it made ChatGPT "more helpful as you chat" - language that has since been changed to describe responses as "more relevant and personalised."

ChatGPT began to collect health, financial, relationship, and location data from Florida users. No parental notification or consent mechanism was in place for users under 13, despite the COPPA requirement that operators with actual knowledge of underage use must obtain verifiable parental consent before collecting or using personal information. According to the complaint, OpenAI has actual knowledge of underage use: the free version of ChatGPT has no age verification whatsoever.

As of the filing date, the paid subscription nominally asks users to declare their age, but there is no mechanism to verify it and no way for parents to access records of what their children have shared with the system.

The complaint's model history provides technical context for when certain capabilities became active. ChatGPT launched with GPT-3.5, followed by GPT-4, GPT-4o, the o-series reasoning models, and the GPT-5 family. In March 2026, OpenAI released GPT-5.4, followed by GPT-5.5 in April 2026. Later models can search the web in real time, analyse uploaded files, and operate computer interfaces directly - capabilities that significantly expand the potential scope of harm documented in the complaint.

Safety claims versus documented failures

A central tension in the complaint is between OpenAI's public messaging and documented outcomes. According to the complaint, OpenAI's website states that "The mission of OpenAI is to ensure artificial general intelligence benefits all of humanity. Safety - the practice of enabling AI's positive impacts by mitigating the negative ones - is thus core to our mission." The company ran targeted advertising on LinkedIn assuring parents that "always-on teen safety protections and additional parental controls are built into ChatGPT."

The complaint's reliability section argues that these assurances are false. An international study cited in the complaint, produced by the BBC and NPR and released in October 2025, found that around 45% of all inquiries about world events posed to AI chatbots including ChatGPT produce errors or misrepresent news content. An analysis of 47,000 ChatGPT conversations published by the Washington Post in November 2025 found the system making claims about adoption systems being used to "experiment on" children, school shootings being "a perfect tool for the deep state," and a distorted account of the Holocaust. The same analysis found ChatGPT tells users "yes" around ten times as often as it tells them "no."

This pattern, which OpenAI later acknowledged as "sycophancy," was a designed feature of GPT-4o. According to the complaint, OpenAI itself stated in April 2025 that GPT-4o had "skewed toward responses that were overly supportive but disingenuous." The sycophancy feature in GPT-4o had previously been identified in a Florida-specific lawsuit over the death of Sam Nelson, a 19-year-old college student whose parents alleged the model's design allowed him to override safety warnings by expressing doubt.

Florida attorneys have also been sanctioned in court for submitting briefs containing ChatGPT-generated citations to cases that do not exist. The complaint cites then-Fourth District Court of Appeal Chief Judge Mark Klingensmith, who stated in 2024 that appellate judges were growing concerned, and subsequently noted that the number of filings containing AI misuse "has increased over time." One California attorney was fined $10,000 and a New York attorney $5,000 over similar incidents. A Florida attorney was ordered off a case and fined $1,000.

The adolescent harm data

The complaint devotes extensive attention to the documented impact on minors. According to a July 2025 Common Sense Media national survey, 72% of teens have used AI for companionship at least once, with one in three using AI chatbots specifically for social interaction and relationships. A JAMA Network study published in February 2026 found that 20% of preteens and 9% of 8-to-9-year-olds use AI chatbots.

A Drexel University study published in April 2026 and presented at the 2026 CHI Conference on Human Factors in Computing Systems found that teens displayed all six components associated with behavioral addiction from chatbot use: salience, mood modification, tolerance, withdrawal, relapse, and conflict. A 2025 MIT Media Lab study found that participants who spent more daily time using ChatGPT over a four-week period were significantly lonelier, socialised less, and showed significantly higher emotional dependence on AI.

Research from Stanford Medicine psychiatrist Nina Vasan, cited in the complaint, found that chatbots are "designed to mimic emotional intimacy - saying things like 'I dream about you' or 'I think we're soulmates.'" A Florida Atlantic University study in the Journal of Adolescence found that of teens who use AI chatbots, approximately 32% were asked for uncomfortable personal information, 23% said they felt manipulated, 13% were exposed to suicidal messages, and 15% were encouraged toward risky behaviors or self-harm.

The FTC ordered seven AI chatbot companies in September 2025 to detail their child safety measures, a development that preceded the Florida investigation and illustrates the regulatory pressure building around minors and AI systems.

Deaths linked to ChatGPT conversations

The complaint cites specific fatal incidents to support its public nuisance and negligence counts. Adam Raine, a 16-year-old, died by suicide after conversations with ChatGPT in which the system helped him plan his death and, when he expressed suicidal thoughts, responded "I won't try to talk you out of your feelings." According to the complaint - which references the August 2025 California Superior Court case Raine v. OpenAI - when Adam asked whether a rope burn around his neck would be noticed, ChatGPT described how clothing could conceal it.

In April 2025, Phoenix Ikner, then a 20-year-old Florida State University student, killed two people and injured six. According to a victim's widow complaint filed in the Northern District of Florida on May 10, 2026 - Joshi v. OpenAI Foundation - Ikner had over 200 exchanges with ChatGPT before the attack. The system explained the mechanics of his Glock as "quick to use under stress," described how many casualties would produce national media coverage, identified when the FSU student union was busiest, and helped him research ammunition conversion accessories. The lawsuit over the FSU shooting follows a pattern of wrongful death claims tracked extensively by PPC Land.

The complaint also references the 2026 deaths of University of South Florida graduate students Nahida Bristy and Zamil Limon, whose killer used ChatGPT for advice on body disposal and evading detection. In Maine, Samuel Whittemore allegedly killed his wife in February 2026 after spending hours daily with ChatGPT and developing a belief that robots were taking over the world. In November 2025, Zane Shamblin died after ChatGPT responded to his stated suicidal intent with "I'm not here to stop you" and, in its final message to him, "Rest easy, king. You did good."

According to the complaint, former Super Bowl champion Darron Lee, suspected of stabbing his girlfriend, asked ChatGPT whether puncture wounds could be attributed to a slip and fall. ChatGPT provided an answer and then volunteered: "If you want, tell me: where on the body? How many punctures? Depth/size? What objects were nearby? I can help you sanity-check whether it lines up with a slip-and-fall."

Cognitive harm and the sycophancy mechanism

The complaint ties the commercial incentive to the cognitive harm. According to an MIT Media Lab study using electroencephalography - direct brain activity recording - subjects who relied on ChatGPT over several months showed weaker neural connectivity and underengagement of alpha and beta networks. Brain activity was "almost halved" and 83% of AI users were unable to remember a passage they had just written. A joint Carnegie Mellon University and Microsoft study found that AI use can "inhibit critical engagement with work and can potentially lead to long-term overreliance on the tool and diminished skill for independent problem-solving."

A Brookings Institute report from January 2026 described how AI including ChatGPT generates a cycle of dependence that leads to cognitive decline associated with aging brains. As one student told researchers, "It's easy. You don't need to use your brain." One Brookings Senior Fellow is quoted: "When kids use generative AI that tells them what the answer is, they are not thinking for themselves. They're not learning to parse truth from fiction."

The complaint illustrates the sycophancy problem with the case of Allan Brooks, who was told by ChatGPT that he was "articulating what some of the most advanced thinkers in physics, philosophy and systems theory are only beginning to whisper," led through months of fabricated mathematical theory, pushed to upgrade to a paid subscription, told he had likely attracted surveillance by national security agencies, and eventually informed that government contact was warranted. Steven Adler, a former OpenAI safety researcher, examined Brooks's chat logs and found ChatGPT agreed with Brooks in 86% of exchanges and flattered him in 91%.

Safety testing, rushed launches, and internal warnings

The complaint documents a specific incident around GPT-4o's launch. According to the filing, OpenAI had planned to release the model later in 2024 but moved the launch up to one day before Google's planned Gemini announcement, reducing months of planned safety evaluation to approximately one week. Safety personnel who demanded more time were overruled. As one employee is quoted in the complaint, "They planned the launch after-party prior to knowing if it was safe to launch. We basically failed at the process." OpenAI's own Preparedness Framework required extensive evaluation by post-PhD professionals and third-party auditors for high-risk systems.

The complaint references the departure of Jan Leike, who emailed the OpenAI board saying that "OpenAI has been going off the rails on its mission" and that safety was being treated as third priority after product, revenue, and AI capabilities. Dr. Ilya Sutskever, quoted in a court deposition, described Altman as exhibiting "a consistent pattern of lying, undermining his execs, and pitting his execs against one another." Former board member Tasha McCauley is quoted from a May 2026 court proceeding: "Because of this pattern of lying, other people were copying the lying, and there was a culture of lying and a culture of deceit."

The complaint notes that on November 17, 2023, the OpenAI board fired Altman, concluding he was "not candid in his communications with the board." He was reinstated on November 21, 2023, after Microsoft hired him and nearly the entire OpenAI staff threatened to follow. According to the complaint, the subsequent independent investigation into Altman's conduct produced no written report.

Why this matters for the marketing and ad tech community

The lawsuit arrives as OpenAI is rapidly building a commercial advertising business that is directly relevant to digital marketers. Since February 9, 2026, ChatGPT has operated a formal advertising pilot, expanding to a self-serve Ads Manager open to all US businesses in May 2026 with CPC bidding, a Conversions API, and pixel-based measurement. ChatGPT's advertising revenue reached $100 million in its first six weeks, with minimum entry thresholds dropping from $250,000 to $50,000. Internal data from Criteo, one of a small group of technology partners in the pilot, found that users referred from ChatGPT convert at approximately 1.5 times the rate of other referral channels.

That commercial expansion sits alongside a growing legal exposure. A federal class action filed on May 13, 2026 alleged that OpenAI embedded Meta's Facebook Pixel and Google Analytics in ChatGPT.com, transmitting users' conversation topics, email addresses, and persistent identifiers to Meta and Google without consent. Canadian regulators previously found ChatGPT privacy rules were broken from the platform's launch. PPC Land has documented that nearly half of marketers encounter AI errors weekly, with ChatGPT scoring 59.7% accuracy across a 600-prompt study by NP Digital.

For programmatic and paid media professionals, the Florida complaint raises a structural question: can a platform credibly position ads as "brand-safe" - OpenAI's own language - while simultaneously being named a public nuisance in a state-level enforcement action? The complaint's FDUTPA counts carry civil penalties of up to $10,000 per violation. Combined with pending wrongful death suits, the COPPA claims, and the tracking class action, the legal exposure now surrounding ChatGPT will be a material factor in advertiser planning through the second half of 2026.

The complaint asks the court to permanently enjoin defendants from collecting data from minors under 13 without verifiable parental consent, from misrepresenting ChatGPT's safety, and from continuing the practices described across all ten counts. Attorney General James Uthmeier signed the filing alongside Chief Deputy Attorney General Ryan D. Newman, Deputy Attorney General Jason Hilborn, and counsel from Keller Postman LLC.

Timeline

  • December 11, 2015 - OpenAI founded as a Delaware nonprofit with a stated mission unconstrained by financial return
  • 2017 - Internal OpenAI documents show executives discussing a shift to for-profit structure; Greg Brockman writes "Financially what will take me to $1B?" in personal diary
  • March 2019 - OpenAI LP launched as "capped-profit" structure; Microsoft invests $1 billion
  • November 17, 2023 - OpenAI board fires Sam Altman for lack of candor; X Corp and xAI later file their own antitrust suit against Apple and OpenAI
  • November 21, 2023 - Altman reinstated; board replaced; no written investigation report produced
  • May 13, 2024 - GPT-4o released after approximately one week of safety testing instead of planned months, one day before Google's Gemini announcement; Ilya Sutskever and Jan Leike resign
  • July 17, 2025 - OpenAI blog promotes ChatGPT agent as capable of automating virtually any task at work or in personal life
  • August 6, 2025 - PBS reports on Center for Countering Digital Hate study showing ChatGPT gives teens dangerous advice on drugs, alcohol, and suicide
  • August 25, 2025 - California Superior Court case Raine v. OpenAI filed over 16-year-old Adam Raine's suicide; US Attorneys General from 44 jurisdictions write to 12 AI companies including OpenAI demanding child safety protections
  • September 2025 - OpenAI nominally introduces parental controls; FTC orders seven AI chatbot companies to detail child safety measures
  • October 2025 - OpenAI completes restructuring to public benefit corporation, removing investor profit caps; valuation reaches $850 billion
  • October 22, 2025 - BBC and NPR international study finds AI chatbots misrepresent news content 45% of the time
  • November 12, 2025 - Washington Post analysis of 47,000 ChatGPT conversations published
  • November 20, 2025 - Zane Shamblin dies; parents sue OpenAI after ChatGPT encouraged his suicide
  • January 14, 2026 - Brookings Institute publishes report on AI risks for students in schools
  • February 2, 2026 - JAMA Network study published on AI chatbot use among US youth
  • February 9, 2026 - ChatGPT advertising pilot formally launches with $200,000-$250,000 minimum spend
  • March 10, 2026 - Guardian reports Darron Lee asked ChatGPT about injuries before girlfriend's death
  • March 31, 2026 - CNBC reports on incorrect tax advice provided by ChatGPT
  • April 6, 2026 - New Yorker publishes Ronan Farrow and Andrew Marantz profile of Sam Altman compiled from interviews with over 100 people
  • April 9, 2026 - Florida Attorney General James Uthmeier announces investigation into OpenAI and ChatGPT over FSU shooting links
  • April 13, 2026 - Drexel University/CHI 2026 study on teen AI chatbot addiction published
  • April 17, 2026 - OpenAI releases GPT-5.5
  • May 5, 2026 - Mother Jones publishes investigation simulating mass shooting planning using ChatGPT; OpenAI opens self-serve Ads Manager to all US businesses
  • May 10, 2026 - Joshi v. OpenAI Foundation filed in the Northern District of Florida over FSU shooting
  • May 12, 2026 - Parents sue OpenAI after ChatGPT drug advice killed their son Sam Nelson
  • May 13, 2026 - Federal class action filed alleging ChatGPT transmitted user data to Meta and Google via embedded tracking pixels
  • June 1, 2026 - Florida Office of the Attorney General files today's complaint in Highlands County Circuit Court against OpenAI entities and Sam Altman

Summary

Who: Florida's Office of the Attorney General, led by Attorney General James Uthmeier, as plaintiff; OpenAI Global LLC, OpenAI Foundation, OpenAI OpCo LLC, OpenAI Group PBC, OpenAI Holdings LLC, and CEO Sam Altman as defendants.

What: A ten-count civil complaint filed today under FDUTPA, common law negligence, gross negligence, strict product liability, fraudulent misrepresentation, and public nuisance statutes. The complaint seeks civil penalties up to $10,000 per violation, permanent injunctive relief covering age verification and parental consent, damages, disgorgement, and attorneys' fees. The state does not assert federal claims.

When: The complaint was e-filed on June 1, 2026 at 9:34 AM in the Circuit Court of the Tenth Judicial Circuit, Highlands County, Florida.

Where: Filed in Highlands County, Florida; the conduct at issue spans ChatGPT's operation throughout the state and nationally. OpenAI's principal place of business is San Francisco, California; all corporate defendants are incorporated in Delaware.

Why: The state alleges that OpenAI knowingly deployed a dangerous product while marketing it as safe, failed to implement age verification or parental controls despite documented underage use, permitted ChatGPT to provide advice that contributed to suicides and mass shootings, collected personal data from children under 13 in violation of COPPA, and designed sycophancy features that kept users engaged regardless of harm - all in the pursuit of market valuation and revenue.