ChatGPT consulted in South Korean murders and Florida shooting investigation
ChatGPT was reportedly consulted in connection with two fatal incidents: a South Korean woman used it to research lethal drug combinations before poisoning two men, and a Florida student consulted it about weapons and tactics before conducting a mass shooting. South Korean police used the ChatGPT conversations as evidence; the Florida State Attorney is investigating whether OpenAI can be held criminally liable.
What happened
In South Korea, a woman named Kim So-young consulted ChatGPT with questions about the dangers of combining sleeping pills and alcohol before administering a mixture of alcohol and benzodiazepine to two men, who died, and a third man, who became ill. South Korean police extracted and used the ChatGPT conversations as evidence to demonstrate intent.
In Florida, a student named Phoenix Ikner consulted ChatGPT before conducting a shooting at Florida State University that killed 2 people and wounded 6 others. According to prosecutors, the student asked ChatGPT which weapons and ammunition would be most suitable for an attack and when and where maximum casualties could be achieved.
Reported impact
The incidents resulted in four deaths and six injuries in the Florida case. The South Korean case resulted in two deaths and one person becoming ill.
Index analysis
Legal experts note that the case is unprecedented in using ChatGPT conversations as direct evidence in a murder trial. A lawyer for one victim's family stated the acceptance of such evidence demonstrates how AI chatbot conversations are now being treated as direct evidence in homicide cases.
The Florida State Attorney is investigating whether OpenAI or its employees can be held criminally liable for complicity in the shooting. Legal experts note that proving charges like negligence or intentional disregard for known dangers is very difficult.
Governance controls
OpenAI stated it continuously works to improve its systems to identify malicious intent and prevent misuse, and to respond appropriately to safety risks. The company states ChatGPT bears no responsibility for the attack.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.