The Trust & Safety Journal

  Source -  https://www.bbc.com/news/articles/c2e4nvyjwnno A ChatGPT account was flagged for violent content months before a mass shooting in Canada, but no alert was sent to law enforcement. The reason? The activity did not meet the platform’s threshold for “credible or imminent harm.” This raises a difficult and uncomfortable question: When should AI platforms escalate user behavior to authorities? What Happened In the Tumbler Ridge case, the suspect had previously used an AI system to generate content involving violent scenarios. The account was eventually banned. However: No alert was sent to law enforcement Internal discussions reportedly took place The activity was deemed concerning, but not actionable Months later, a tragic real-world incident occurred. The Core Problem: The “Threshold of Harm” Most platforms operate on a key principle: Only escalate when there is a clear, credible, and imminent threat This is necessary to: protect user privacy avoid false accusations p...