'Could it kill someone?' A Seoul woman allegedly used ChatGPT to carry out two murders in South Korean motels

South Korean authorities have uncovered what appears to be one of the first documented cases of AI-assisted homicide, with digital forensics revealing a suspect's extensive use of ChatGPT to research lethal pharmacology. During the investigation into the Gangbuk Motel Serial Deaths, police discovered that a 21-year-old woman repeatedly prompted OpenAI's conversational AI to understand the effects of mixing benzodiazepine-class sleeping pills with alcohol, explicitly asking whether such combinations could cause death. The case represents a chilling escalation in how generative AI tools can be weaponized, moving beyond misinformation campaigns into direct physical harm. Prosecutors are now pursuing elevated murder charges based on evidence that the suspect systematically used the AI as a research tool before allegedly administering fatal overdoses to multiple men in Seoul motels.

Forensic analysis showed the suspect continued her queries even after ChatGPT provided clear warnings about the potentially fatal consequences of mixing the drugs with alcohol. According to investigation documents, she allegedly proceeded to double the drug dosage on her victims—a deliberate escalation that resulted in two confirmed deaths and left a third victim in a coma. The case has triggered urgent discussions among South Korean legal experts about establishing new precedents for AI-facilitated crimes and whether existing laws adequately address this novel form of premeditation. As AI systems like GPT-4 become more capable of providing detailed, accurate information across domains including chemistry and medicine, this incident highlights the urgent need for both technical safeguards and legal frameworks to prevent malicious use while preserving legitimate access.