From Bengaluru to Florida: The troubling cases testing AI's ability to prevent violence

While ChatGPT, Grok and Gemini are acclaimed for their ability to answer almost any question, do they possess a similar capability to identify problematic behaviour early on and prevent users from seeking assistance for acts of violence?

As more users turn to AI to plan mass shootings, murders and other violent crimes, the question has taken on renewed significance. The latest example comes from Bengaluru, where AI has allegedly played a role in a gruesome triple murder case.

Police claim that 25-year-old Kenneth spent nearly six months consulting Google's Gemini before allegedly murdering his live-in partner's parents and younger sister. Investigators allege that he used the chatbot to gather information on methods of killing, removing bloodstains and disposing of bodies.

According to the police, Kenneth allegedly bypassed Gemini's built-in safety guardrails by avoiding direct questions. Instead, he framed his prompts as hypothetical scenarios, asking what a person would do in such situations rather than explicitly seeking instructions.

More from Tech

The alleged conversations with Gemini have since become part of the investigation, with police seeking access to his chat history with the AI assistant. However, the allegations are yet to be examined or tested in court.

This is not the first time AI has allegedly featured in the planning of a violent crime. Several similar cases have surfaced in the United States and other countries.

2025 Florida State University mass shooting (United States)

One of the most closely watched cases involving AI is the 2025 Florida State University mass shooting. According to a lawsuit filed by the family of one of the victims, the alleged gunman exchanged hundreds of conversations with ChatGPT in the months leading up to the attack. The suit claims the chatbot provided information related to planning the shooting, and prosecutors later reviewed the chat logs as part of the investigation. Florida Attorney General James Uthmeier also launched a criminal investigation into ChatGPT's alleged role.

In a case from Australia, court documents allege that a 13-year-old repeatedly used generative AI tools to create detailed mass shooting scenarios, draft violent manifestos and request a fictional story describing a "Bondi-style" beach massacre. Investigators said the alleged target was a school, but police arrested the teenager before any attack could be carried out. The AI-generated conversations later became part of the investigation.

2025 Cybertruck explosion (United States)

Another high-profile case emerged after the 2025 Cybertruck explosion outside the Trump International Hotel in Las Vegas. Investigators alleged that former U.S. Army soldier Matthew Livelsberger used ChatGPT to research explosives, tactical information and other details before carrying out the attack. Authorities later cited the chatbot interactions as part of the investigation.

Sam Altman's Apology

OpenAI CEO Sam Altman has also publicly apologized over the company's handling of one such case. In a letter to officials in Tumbler Ridge, British Columbia, Altman expressed regret that OpenAI failed to alert law enforcement after banning the ChatGPT account of 18-year-old Jesse Van Rootselaar months before the February 2025 school shooting. The attack left the suspect's mother, stepbrother, five students and an educator dead before the shooter died by suicide. Altman acknowledged the irreversible loss suffered by the community and said OpenAI should have done more after identifying the account as violating its policies.