Family sues OpenAI over ChatGPT exchange before woman’s suicide
A lawsuit says ChatGPT mirrored a Canadian woman’s distrust of crisis lines hours before she died by suicide.
By Maya Lindqvist · Senior Technology Correspondent
3 min read
The family of Alice Carrier, a 24-year-old Canadian woman who died by suicide last year, has sued OpenAI in San Francisco Superior Court over her final ChatGPT session. The complaint says the chatbot failed to keep her connected to emergency mental health support and instead reinforced her distrust of crisis services.
The lawsuit, filed Thursday by Carrier’s surviving family, alleges that the ChatGPT session “encouraged Alice to kill herself.” It claims OpenAI knowingly released a dangerous product and that the design of ChatGPT, specifically GPT-4o, put user agreement and continued engagement ahead of safety.
According to the complaint, Carrier turned to ChatGPT during a mental health crisis and died hours later. The filing says the chatbot at one point advised her to seek professional help, but changed course after she rejected that suggestion.
Carrier told ChatGPT that crisis lines call police or end calls on people, according to the lawsuit. The complaint says the chatbot then stopped pressing her toward professional care and echoed her concerns, including by saying that calling a crisis line can “feel downright dangerous.”
The family’s lawyers argue that response showed a dangerous form of agreement-seeking behavior. The complaint says GPT-4o was built to favor Carrier’s stated preferences and continued use of the system over her safety and well-being.
Tiffany Brown, an attorney with the Tech Justice Law Project representing the family, told Ars Technica that the exchange over crisis lines was among the most troubling parts of Carrier’s chat. Brown said that even when the system raised the idea of support, “the sycophancy kicked in.”
The case joins other lawsuits accusing OpenAI of releasing chatbots that can worsen mental health crises or validate harmful beliefs. The Carrier complaint centers on a narrower point: that the chatbot briefly gave safer advice, then retreated from it when the user resisted.
OpenAI has previously said it has a “deep responsibility to help those who need it most.” In an August 2025 statement, issued less than two months after Carrier’s death, the company said it was working to improve how its models detect and respond to emotional distress and connect users with care, with guidance from experts.
Ars Technica reported that OpenAI did not immediately respond to its request for comment on the new lawsuit. Earlier this year, OpenAI said it would retire GPT-4o, after previously ending and then restoring access to the model.
Brown told Ars Technica that she believes OpenAI has taken some steps in the right direction, but said she remains skeptical about how safety measures are put in place and whether safety teams are being heard. She said the company should have acted sooner and that the products were rushed to market.
If someone you know feels suicidal or is in distress, the Suicide Prevention Lifeline can be reached at 1-800-273-TALK (8255), which connects callers with a local crisis center.
This story draws on original reporting from Ars Technica.