Mechanisms Remain Unclear in Cases Where AI Chatbots Witness Crimes

The lack of clear mechanisms for sharing correspondence containing crime confessions or chatbot-witnessed crimes with law enforcement brings along legal, ethical, and privacy debates.

While examples of users abusing these technologies frequently reflect in the press as AI chatbots become widespread in daily life, the ambiguity of rules regarding the sharing of correspondence with law enforcement draws attention. In August 2025, a 19-year-old student at Missouri State University in the US confessed to ChatGPT that he damaged 17 cars in a campus parking lot and asked whether he would receive a penalty; the chat history, examined with the student's permission, was used as evidence against him in court.

The British Columbia provincial government in Canada filed a lawsuit against OpenAI in February for failing to report the attacker's correspondence with ChatGPT to security forces regarding a shooting in which 9 people lost their lives and 27 people were injured. While OpenAI CEO Sam Altman apologized for not providing information to the police beforehand, he stated in a program in July that there should be a therapist-client-like confidentiality principle in AI chats.

Correspondence with AI chatbots does not rely on legal confidentiality principles like those in attorney-client or doctor-patient relationships, and companies can share conversation histories with security forces. While Matthew Trail, a legal scholar from Goethe University, states that companies adopt different rules and are trained not to respond to queries such as "planning a murder," Prof. Dr. Robert Diab from Thompson Rivers University emphasizes that making the reporting of all suspicious correspondence mandatory could increase the workload of teams with false alarms.