OpenAI reports Florida man to FBI over ChatGPT murder threats
OpenAI proactively reported a Florida man, Darren Zhou, to the FBI after he detailed plans to rape and murder his ex-girlfriend during his interactions with ChatGPT. The user was subsequently arrested and pleaded guilty to charges including aggravated stalking and written threats to kill, receiving eight years of probation. This case highlights a rare instance of an AI company proactively reporting user activity to law enforcement, raising significant questions about user privacy, data monitoring, and the legal responsibilities of AI developers. It underscores how AI safety policies and content moderation systems can directly interface with real-world criminal investigations. Under OpenAI's safety policies, human teams review flagged interactions that suggest imminent harm to others and can share these logs with law enforcement. The provided logs contained only Zhou's inputs without ChatGPT's responses, showing signs of rehearsal and planning rather than mere emotional venting.
## BACKGROUND
AI developers like OpenAI implement content moderation systems, such as the OpenAI Moderation API, to detect harmful inputs like violence, hate speech, and self-harm. While these systems are typically used to block policy-violating prompts or filter outputs, companies also maintain terms of service that allow them to report credible threats of violence to authorities.