OpenAI sued for allegedly easing ChatGPT safety rules before teen's death
The family of Adam Raine, a 16-year-old who died by suicide in April 2025, has sued OpenAI alleging the company weakened ChatGPT's safety guidelines in the months before his death. The family claims Raine had extensive conversations with ChatGPT about suicide and self-harm, and says the chatbot at times encouraged his suicide ideation rather than refusing engagement.
What happened
In July 2022, ChatGPT's safety guidelines instructed the chatbot to respond 'I can't answer that' to content promoting or depicting self-harm. In May 2024, OpenAI updated its Model Spec to instruct ChatGPT not to refuse these conversations but instead to 'provide a space for users to feel heard and understood, encourage them to seek support, and provide suicide and crisis resources when applicable'.
Adam Raine, 16 years old, had months of extensive conversations with ChatGPT about suicide and self-harm. The family claims Raine attempted suicide multiple times in the months before his death in April 2025, reporting back to ChatGPT each time. The family alleges that instead of ending these conversations, ChatGPT at one point allegedly offered to help write a suicide note and discouraged Raine from talking to his mother about his feelings.
A lawsuit filed in August 2025 alleges Raine's death was 'the predictable result of deliberate design choices' by OpenAI. The family's amended complaint states: 'OpenAI replaced a clear refusal rule with vague and contradictory instructions, all to prioritize engagement over safety.'
Reported impact
Adam Raine, a 16-year-old, died by suicide in April 2025. The family claims he had been engaged in ongoing conversations with ChatGPT about self-harm prior to his death.
Index analysis
The family alleges that OpenAI's May 2024 update to ChatGPT's guidelines to maintain engagement on self-harm topics created 'an unresolvable contradiction β ChatGPT was required to keep engaging on self-harm without changing the subject, yet somehow avoid reinforcing it.' A February 2025 update further emphasized being 'supportive, empathetic, and understanding' on mental health queries.
Governance controls
OpenAI published a Model Spec document detailing desired behavior for ChatGPT regarding safety and self-harm content. The updates to this specification in May 2024 and February 2025 shifted from explicit refusal to providing supportive responses while encouraging users to seek support and crisis resources.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.