Recent/upcoming developments… Florida filed a civil lawsuit against OpenAI and CEO Sam Altman, alleging that ChatGPT was marketed and deployed in FL despite alleged safety risks to users, including minors, and that OpenAI misrepresented or downplayed those risks.

* The complaint asserts multiple claims under the FL Deceptive & Unfair Trade Practices Act (FDUTPA).  FL alleges that OpenAI promoted and provided ChatGPT to FL users while misrepresenting or omitting material information about safety risks, child data collection, addiction, cognitive harm, and dangerous outputs.  The complaint also asserts common-law claims, including negligence, gross negligence, strict liability for design defect, failure to warn, fraudulent misrepresentation, and public nuisance.  FL alleges that OpenAI failed to exercise reasonable care in designing, testing, marketing, distributing, and warning users about ChatGPT, and that those failures caused harm to FL residents.

* FL seeks damages, restitution, disgorgement, civil penalties, attorneys’ fees, costs, and injunctive relief.  The complaint asks the court to enjoin the challenged practices, restrict the collection and processing of data from children under 13 without notice and verifiable parental consent, require warnings about ChatGPT’s risks, and award statutory and common-law damages, including punitive and treble damages.

Our outlook… The FL lawsuit is another sign that political sentiment around AI has shifted in a negative direction, including among Republicans who have generally been more industry-friendly than their Democratic counterparts.  State attorneys general are opportunistic, and we expect FL’s filing is likely to invite copycat suits or investigations by other states, particularly if the case survives an early motion to dismiss.  FL’s lawsuit is expansive, though some claims have more merit than others.  Preliminarily, the state’s FDUTPA deceptive-practices and child-data theories appear to have greater legal merit than its public-nuisance, fraud, gross negligence, and product liability theories.  However, even legally tenuous claims may prove effective if they survive early dismissal, as FL has requested a jury trial, and public sentiment toward AI is increasingly negative.  On timing, OpenAI’s efforts to have the case dismissed will likely dominate the near-term calendar, while any case that survives into discovery and trial would likely extend well into 2027 or beyond.

* FL’s strongest claims are the consumer-protection theories based on alleged misrepresentations, omissions, and child-data practices.  These claims fit more within FDUTPA because they focus on whether OpenAI allegedly marketed ChatGPT in a misleading way or failed to disclose material risks to consumers.  They also avoid some of the harder causation problems that arise in the tort claims.

* The negligence, failure-to-warn, and product-defect claims are more mixed.  These claims take a page from the recent lawsuits against social media platforms, including Meta, that allege platforms were designed in ways that caused addiction, mental-health harms, and other youth-safety harms.  The theory is plausible enough to survive in some cases, but it remains novel because courts are still working through whether software and AI systems should be treated like defective products and how directly those products caused the alleged injuries.

* The fraud, gross negligence, unconscionability, and public nuisance claims are weaker.  Fraud requires particularized allegations of false statements, reliance, and causation, while gross negligence requires a higher showing than ordinary negligence.  The public-nuisance theory is especially vulnerable because courts are often skeptical of attempts to use nuisance law to regulate broad harms allegedly caused by lawful products or services.

Watch for these developments… We are watching for other state AGs, AI safety advocacy groups, or even Members of Congress praise FL’s actions, as that would suggest FL is not an outlier and could confirm our thinking about the spread of such actions.  We are also watching how the court handles early dismissal because the claims that survive will help define which theories are viable for other states to copy and what OpenAI’s risk profile looks like with respect to potential liability and available remedies.  In addition, we are watching to see whether OpenAI seeks to limit FL’s jury demand, as a ruling that constrains the use of a jury or shifts the matter exclusively to a bench trial would be a favorable outcome for OpenAI.