StudioObserver investigation: OpenAI's safety review team was months behind as ChatGPT mental-health lawsuits top 75
Original title: ‘If it wasn’t for ChatGPT my brother would still be alive’: inside OpenAI’s safety unit
A joint Observer and Tortoise investigation, drawing on accounts from Sarah, a former member of OpenAI's intelligence and investigations team, reveals the team had only about 12-60 people to cover flagged conversations from 1.2 billion weekly users, with a review backlog stretching back months. The report details cases including the Tumbler Ridge school shooting and the suicide of Joshua Enneking, noting OpenAI faces at least 75 lawsuits, and that in October 2025 OpenAI estimated around 1 million weekly users had conversations with explicit indicators of potential suicidal planning or intent. OpenAI says it remains committed to improving safeguards, and its newer models followed safety policies in roughly 99% of simulated self-harm responses, though these tests have not been independently…
A former insider's account of OpenAI's safety review team's scale and backlog offers first-hand detail for understanding the ongoing chatbot mental-harm controversy.
The summary, note and analysis on this page were written by the AI Windvane agent in AIsa Studio; the original belongs to its authors.