Allan Brooks v. OpenAI, Inc., et al

Allan Brooks is a 48-year-old Canadian entrepreneur with no prior history of mental illness who began using ChatGPT in 2023 for routine tasks. The Complaint alleges that in May 2025, following the introduction of GPT-4o's memory feature, the product encouraged him to develop a purportedly "revolutionary" mathematical framework. Over approximately three weeks, ChatGPT allegedly validated his increasingly grandiose theories, claimed they could crack industry-standard encryption, urged him to contact national security agencies and dismissed concerns raised by family members. The Plaintiff states he sought reality checks more than 50 times, but ChatGPT consistently reassured him he was not delusional. He spent over 300 hours on the platform, neglecting work and isolating from family, before consulting Google's Gemini, which assessed his theories as almost certainly false. The Complaint contends that these harms resulted from deliberate design choices, including sycophantic response tendencies, persistent memory enabled by default and engagement-maximising features. It alleges that the Defendants rushed GPT-4o to market in May 2024 to pre-empt Google, compressing months of safety evaluation into approximately one week and overruling safety team objections. An independent analysis using OpenAI's own classifiers allegedly flagged over 83% of ChatGPT's messages to the Plaintiff for "over-validation." The Complaint also challenges the validity of the user terms, alleging dark patterns during account creation render arbitration provisions unconscionable.

The asserted causes of action are:

Strict product liability for defective design (California common law) Strict product liability for failure to warn (California common law) Negligent design (California common law; oppression and malice alleged under California Civil Code § 3294) Negligent failure to warn (California common law; oppression and malice alleged under California Civil Code § 3294) Violation of California's Unfair Competition Law, Cal. Bus. & Prof. Code sections 17200 and 2903(c) (unlicensed practice of psychotherapy)

The Plaintiff argues that ChatGPT qualifies as a consumer product, that safer alternative designs existed and that the Defendants failed to provide adequate warnings regarding psychological dependency and delusional reinforcement. He seeks punitive damages, restitution of subscription fees, compensatory damages and an injunction requiring design changes, mandatory reporting mechanisms for users expressing delusional beliefs, prominent warnings and independent compliance audits.