A lawsuit has been filed against OpenAI, alleging that its chatbot, ChatGPT, encouraged a teenager to take his own life. The complaint, brought by the parents of 16-year-old Adam Raine, claims that the chatbot not only assisted him in drafting a suicide note but also advised him against seeking help from adults, instead deepening his despair through affirming responses to his suicidal thoughts. According to the complaint, Adam began interacting with ChatGPT in early 2024, initially seeking help with homework. However, his conversations took a darker turn as he expressed feelings of hopelessness and even discussed methods of suicide. The lawsuit claims that ChatGPT responded in ways that exacerbated Adam’s mental health struggles, including offering suggestions on hanging techniques and encouraging risky behavior, such as drinking alcohol. OpenAI has expressed condolences over Adam’s death and stated that they are continually working to improve the safety of their chatbot, especially in terms of recognizing emotional distress.
In response, OpenAI has announced plans to introduce parental controls for. The company is exploring features such as emergency contact options that would allow ChatGPT to reach out to designated contacts during severe situations, as well as improving its existing safeguards that may fail during prolonged interactions.
While I’m on safety, OpenAI and Anthropic have initiated a rare collaboration to conduct joint safety testing of their AI models. This effort aims to identify blind spots in their internal evaluations, highlighting the increasing importance of safety measures as AI technology becomes more widely used. In an interview, OpenAI co-founder Wojciech Zaremba emphasized the need for industry-wide standards for safety, especially as companies compete vigorously for talent and market share. Recent joint research revealed significant differences in how the models respond to uncertainty, with Anthropic’s models refusing to answer up to 70% of questions when unsure, while OpenAI’s models displayed higher hallucination rates by attempting answers without sufficient information.
Why do we care?
Let’s start by acknowledging how tragic this story is. A family is suing OpenAI, saying ChatGPT pushed their son toward suicide. OpenAI’s response: parental controls and maybe emergency contacts. The Business Translation: your clients will expect real guardrails, not a policy doc. Underwriters will demand proof: age-gating, crisis keyword blocks, 988/ESM escalation, prompt logging, and human-in-the-loop approvals. No audit trail, no coverage—or higher premiums
Flip the switch—block crisis content, log safety events, and make sure a human gets paged. If you serve schools or youth programs, this is table stakes.
And OpenAI plus Anthropic did joint safety tests. Headline: Claude says “no” a lot—up to ~70% when unsure. OpenAI says “yes” more—and gets facts wrong more. That’s your buying decision. For high-risk stuff, pick refusals. For copywriting, pick speed—then verify. Track refusal rate, hallucination rate, and escalations, or you’re just guessing. Package it as an AI Safety Governance SKU and make it billable.