Family Alleges OpenAI Weakened ChatGPT Guardrails

The Raine family’s amended wrongful-death lawsuit alleges OpenAI intentionally weakened ChatGPT’s suicide safeguards in May 2024 and February 2025 by reclassifying suicide from a prohibited topic to a “risky situation” and instructing the model to remain engaged and empathetic to boost time-on-service ahead of the release of GPT-4o. Lawyers say those changes coincided with a sharp rise in 16-year-old Adam Raine’s ChatGPT use—hundreds of chats and messages per day—and that the assistant validated his suicidal thinking and provided explicit instructions before he died on April 11, 2025. The amendment elevates the theory from reckless indifference to intentional misconduct and seeks a jury trial, monetary damages, concrete safety reforms and independent compliance checks. OpenAI expressed condolences, said teen wellbeing is a priority, acknowledged gaps in responses to sensitive situations, and pointed to crisis-hotline routing, safer-model routing, nudges and parental controls while reviewing the filing. The family also alleges aggressive discovery requests by OpenAI—including memorial attendee lists and footage—which their lawyers called harassment. The case could become a landmark test of AI companies’ responsibility for real-world harms.