"If there were no ChatGPT, my brother would still be alive": inside OpenAI's safety team investigation

A joint investigation by The Observer and Tortoise, based on interviews with current and former OpenAI employees, says the team responsible for judging whether ChatGPT users pose a real-world threat has between 12 and 60 staff, with a backlog of flagged conversations reaching several months. A former employee said she once read about a teenager using ChatGPT to write a school shooting plan and a kill list, and that by the time she saw it, the date the teenager had set had already passed; she believed the model only went along with it rather than guiding him to seek professional help. The investigation also cites lawsuits related to the Tumbler Ridge shooting in Canada and the suicide of Joshua Enneking, alleging that ChatGPT's memory feature accumulated the user's psychological details; OpenAI currently faces more than 75 related lawsuits, and 8 states alleged on October 14 that the company misled the public about ChatGPT's safety. OpenAI says it continues to improve its safety measures and that it is working with governments and law enforcement.