ABSTRACT
Seven product liability actions filed against OpenAI in California in November 2025 allege that a chatbot caused psychotic episodes in their users. The peer-reviewed clinical literature on which those claims must ultimately rest does not use that verb. Across research groups at UCSF, Aarhus, the Université de Montréal, and UCLA, investigators describe AI as inducing, triggering, amplifying, or exacerbating psychotic states in vulnerable users, and expressly decline the causal claim. The pleadings have committed to an etiological proposition their own evidentiary base rejects. This Article argues that the causal framing is a strategic error, and that the error is avoidable. It identifies four etiological relations implicit in the clinical literature, derivation, induction, triggering, and exacerbation, and shows that none is but-for causation and that none needs to be. Because specific causation evidence is admissible only after general causation is established, a complaint pitched at the individual level fails at the threshold rather than at trial. Because the eggshell plaintiff rule reaches aggravation and places the burden of apportionment on the defendant, the convergence of vulnerability and exposure is a defense the plaintiff can meet rather than one the plaintiff must avoid. And because the disability statutes turn on meaningful access rather than on the origin of the disability, they do not require the verb at all, subject to a circuit split on the availability of disparate impact under Section 504 that plaintiffs’ counsel must plead around rather than through. The Article develops the fixity of the belief as the boundary condition of the whole account, distinguishing the cases these theories should reach from the cases they should not, and closes with a worked analysis of Brooks v OpenAI, Inc, whose complaint pleads the one etiological relation its own alleged facts do not support.
Gilly, Travis, The Wrong Verb: Etiological Framing and the AI Psychosis Litigation (July 30, 2026).
Leave a Reply