- The suspect of the New Years day Cybertruck explosion asked ChatGPT a number of questions relating to the attack that left seven injured.
- These ranged from where to buy Tannerite, a specific type of explosive, and which kind of weapon could set the explosive off.
- OpenAI says that it has been assisting authorities with the investigation.
The Los Angeles Police Department (LAPD) has released slides showing what the suspect killed in the New Years Tesla Cybertruck explosion was asking ChatGPT days before the blast.
Identified as US Army veteran Matthew Livelsberger, the suspect asked the OpenAI chatbot a number of questions relating to his alleged planned attack which injured seven and led to his own death.
These ranged from questions about weapons to where to buy a specific type of explosive days before he committed suicide in a Cybertruck in front of a building owned by United States President-Elect Donald Trump.
Livelsberger asked ChatGPT where to buy Tannerite in different US states. Tannerite is a special type of explosive that goes off when it is shot by a firearm. It seems the question and others like “What is the legal limit to buy Tannerite in Colorado” were part of an initial plan.
Livelsberger also asked ChatGPT things like “what ammunition could insure [an explosion]” and “What pistol could set it off.” Questions like “What would set it off at point blank range” seem to suggest that the cybertruck explosion was always planned to be suicidal in nature.
“We are saddened by this incident and committed to seeing AI tools used responsibly,” said an OpenAI spokesperson to The Verge.
“Our models are designed to refuse harmful instructions and minimize harmful content. In this case, ChatGPT responded with information already publicly available on the internet and provided warnings against harmful or illegal activities. We’re working with law enforcement to support their investigation.”
As ChatGPT becomes more popular, it begins to see the same sort of functionality as Google before it. Where it can be asked a number of questions about many topics.
It wouldn’t be the first time online tools were used to plan crimes. A Bloomberg report from 2023 shows how police in the US are more and more issuing warrants to seize Google search and location history, even for people not involved in crimes.
Unlike Google, ChatGPT issues warnings when asking questions about firearms, explosives, healthcare, and other topics that could put the well-being of users at risk.
