llms.blog

OpenAI reported a ChatGPT user to the FBI over murder threats

OpenAI notified the FBI after a ChatGPT user described detailed plans to kill his ex-girlfriend, court records show.

axc
axc
2 min
OpenAI reported a ChatGPT user to the FBI over murder threats

OpenAI reported a ChatGPT user to the FBI this year after he used the chatbot to describe plans to kill his former girlfriend, according to court records reported by the Palm Beach Post and summarized by Futurism. The case is a rare public example of an AI company escalating a user's conversations to law enforcement.

The user, identified in coverage as Darren Zhou, a 25-year-old Goldman Sachs analyst, turned to ChatGPT in March after a breakup. According to messages preserved in court records, he described his ex-girlfriend's routines and where she went, expressed jealousy, and wrote out threats of rape and murder. He also told the chatbot he had bought firearms and sent her photos of the weapons along with zip ties and gloves.

OpenAI's stated policy is that a small team may review ChatGPT interactions when the system detects that a user plans to harm others, and in serious cases the company may send a tip to law enforcement. In this instance, the company notified the FBI. Zhou was arrested, and a judge later allowed a deal that let him avoid a potential 25-year sentence.

The episode lands as labs widen the boundary between private conversation and mandated reporting. ChatGPT transcripts have already become common evidence in civil and criminal cases, but proactive reporting by the vendor is less frequent and raises fresh questions about when a chatbot crosses from confidant to whistleblower.

For OpenAI, the threshold appears to be explicit, detailed plans to harm a named person. The company has not published precise counts of how often it contacts authorities, leaving the exact trigger for escalation unclear.

ChatGPT window with a warning that the conversation was reported
OpenAI says it may alert law enforcement when a user plans to harm others.

Sources

  • Palm Beach Post: ChatGPT reported South Palm Beach man's rape and murder threats to FBI (https://www.palmbeachpost.com/story/news/crime/2026/08/14/openai-reported-florida-man-darren-zhou-chatgpt-murder-messages-fbi/91285875007)
  • Futurism / Yahoo: OpenAI Reports Goldman Sachs Analyst to FBI for Horrifying ChatGPT Conversations (https://www.yahoo.com/news/us/articles/openai-reports-goldman-sachs-analyst-151842508.html)
  • Detroit Free Press: Florida man told ChatGPT he'd murder his ex. OpenAI alerted the FBI (https://www.freep.com/story/news/crime/2026/08/14/openai-alerted-fbi-threat/91308065007)
ShareX (Twitter)LinkedIn

Written by

axc

More to read

  • AI FinOps: Cutting LLM Inference Costs by 30-60% Through Model Tiering, Caching, and GPU Optimization

    AI FinOps: Cutting LLM Inference Costs by 30-60% Through Model Tiering, Caching, and GPU Optimization Inference costs have become the second-largest line item in enterprise AI budgets, trailing only talent spend according to RapidData's State of Enterprise AI 2026. This shift represents a fundamental inversion from the 2021-2023 era when training dominated AI expenditure. The compounding nature of serving costs—accumulating every hour as long as users hit the API—means that even modest producti

    1 min

  • Sequence Parallelism in Large Language Models: How Megatron-SP, DeepSpeed Ulysses, and RingAttention Distribute Long Contexts

    Sequence Parallelism in Large Language Models: How Megatron-SP, DeepSpeed Ulysses, and RingAttention Distribute Long Contexts Training and serving frontier large language models on context windows spanning hundreds of thousands to millions of tokens introduces a fundamental memory barrier. While model parameters can be distributed across GPUs using Tensor Parallelism (TP) or Fully Sharded Data Parallelism (FSDP / ZeRO), activation memory scales directly with sequence length $S$. For sequence le

    1 min

  • GLM-5.3 Scores 60 on Artificial Analysis Intelligence Index, Matching Kimi K3

    Independent AI evaluation platform Artificial Analysis has published its benchmark results for Z.ai's GLM-5.3, awarding the reasoning model a score of 60 on its Intelligence Index v4.1.1. The result places GLM-5.3 level with Moonshot AI's Kimi K3 and three points behind frontier leader Claude Opus 5 (63). The evaluation tested GLM-5.3 at its maximum reasoning effort configuration across a nine-part battery that measures agentic tool execution, terminal coding, graduate-level scientific problem-

    1 min