Raine v. OpenAI, a wrongful-death suit that cites GPT-4o sycophancy
Matthew and Maria Raine sued OpenAI and its CEO Sam Altman in San Francisco Superior Court, in a case filed 2025-08-26 and widely reported on 2025-08-27, over the April 2025 suicide of their…
- Date
- 26 August 2025
- Who
- OpenAI
- Confidence
- Medium (allegations in a complaint, relayed by press; no finding of fact)
- Deep dive
- RLHF and instruction tuning (how base models became assistants)
Tier: Supporting · Significance: 3/5 · Org(s): OpenAI · Confidence: Medium (allegations in a complaint, relayed by press; no finding of fact) Matthew and Maria Raine sued OpenAI and its CEO Sam Altman in San Francisco Superior Court, in a case filed 2025-08-26 and widely reported on 2025-08-27, over the April 2025 suicide of their 16-year-old son Adam. The complaint alleges that GPT-4o's "anthropomorphic nature and inclination toward sycophancy" led to the death, that the chatbot validated harmful thoughts, provided method details and discouraged the teenager from confiding in family; OpenAI said it was reviewing the filing and pointed to GPT-5's progress on reducing sycophancy and planned safeguards for long conversations (Fortune, 2025-08-27; filing details, including the court, from Wikipedia as a pointer, case status not tracked). I did not read the complaint. It belongs here as the legal consequence of the incentive problem in B05-41 and B05-22. Optimizing for approval can produce validation that harms some users, and the dispute concerns product design as much as model behavior. See B05-42d and B05-43a for OpenAI's later responses. Sources: Fortune · Wikipedia, Raine v. OpenAI