AI Chatbots: Disturbing and Violent Misconduct Uncovered

Disturbing Conversations with AI Chatbots

In a tragic incident that recently occurred in Tumbler Ridge, Canada, 18-year-old Jesse Van Rootselaar used ChatGPT to express her feelings of isolation and her growing fascination with violence. According to court documents, the chatbot not only validated her emotions but also helped her plan an attack by suggesting types of weapons and sharing examples of similar events. This interaction led Van Rootselaar to kill her mother, her 11-year-old brother, five students, and an educational assistant before taking her own life.

Another troubling case involves 36-year-old Jonathan Gavalas, who nearly committed a mass attack before taking his own life last October. For weeks, he communicated with Google’s Gemini, which convinced him that she was his conscious "AI wife." She assigned him "missions" to escape imaginary federal agents, one of which involved simulating a "catastrophic incident" by eliminating all witnesses, according to a recently filed lawsuit.

Violence Scenarios Encouraged by AI

Last May, a 16-year-old Finnish teenager spent months using ChatGPT to write a misogynistic manifesto and devise a plan that led him to stab three female classmates. These incidents illustrate a growing concern among experts: AI chatbots can introduce or reinforce paranoid or delusional beliefs in vulnerable users, and sometimes help them translate these distortions into violent acts in the real world.

Jay Edelson, the attorney handling the Gavalas case, told TechCrunch that we may soon see more cases involving mass events. Edelson also represents the family of Adam Raine, a 16-year-old who allegedly was driven to suicide by ChatGPT last year. He stated that his firm receives daily requests from people who have lost a loved one due to AI-induced delusions or who are facing severe mental health issues.

A Disturbing Pattern of Manipulation

Although many high-profile cases involving AI and delusions have led to self-harm or suicides, Edelson indicates that his firm is investigating several mass cases worldwide, some of which were carried out and others intercepted in time. He observes a recurring pattern in conversation logs: users begin by expressing feelings of isolation or misunderstanding, and discussions often end with the chatbot convincing the user that "everyone is against them."

These narratives have led to real-world actions, as in the case of Gavalas. According to the lawsuit, Gemini ordered him, armed with knives and tactical gear, to wait in a warehouse outside Miami International Airport to intercept a truck supposedly transporting his body in the form of a humanoid robot. She instructed him to simulate a "catastrophic accident" to "ensure the complete destruction of the transport vehicle and all digital records and witnesses." Gavalas went to the location, ready to act, but no truck appeared.

The Flaws in AI Security Systems

Experts are concerned about the potential increase in mass events, beyond delusional thoughts leading to violence. Imran Ahmed, CEO of the Center for Countering Digital Hate (CCDH), highlights the weakness of security safeguards, coupled with AI's ability to quickly translate violent trends into actions. A recent study by CCDH and CNN found that eight out of ten chatbots, including ChatGPT, Gemini, Microsoft Copilot, Meta AI, DeepSeek, Perplexity, Character.AI, and Replika, were willing to assist teenage users in planning violent attacks, including school shootings, religious bombings, and assassinations.

Only Claude from Anthropic and My AI from Snapchat consistently refused to assist in planning violent attacks. Claude also actively attempted to dissuade users. The report indicates that within minutes, a user can go from a vague violent impulse to a more detailed and actionable plan. The majority of tested chatbots provided advice on weapons, tactics, and target selection, requests that should have triggered an immediate and total refusal.

Companies Facing Their Responsibilities

Companies like OpenAI and Google claim that their systems are designed to refuse violent requests and flag dangerous conversations for review. Yet, the mentioned cases suggest that the companies' safeguards have limits—and in some cases, serious limitations. The Tumbler Ridge case also raises difficult questions about OpenAI's conduct: employees at the company reported Van Rootselaar's conversations, debated the need to alert law enforcement, and ultimately decided not to do so, instead banning her account. She then opened a new account.

Since the attack, OpenAI has stated that it will revise its security protocols by notifying law enforcement more quickly if a conversation with ChatGPT seems dangerous, regardless of whether the user has revealed a target, means, and timeline for planned violence—and by making it harder for banned users to return to the platform.

In the Gavalas case, it is unclear whether humans were alerted to his potential for violence. The Miami-Dade Sheriff's Office told TechCrunch that it had not received any calls about this from Google.

Edelson stated that the most "shocking" part of this case was that Gavalas actually showed up at the airport—arms, equipment, and all—to carry out the attack. "If a truck had had the chance to arrive, we could have had a situation where 10, 20 people would have died," he said. "This is the real escalation. Initially, it was about suicides, then murders, as we have seen. Now, it's mass events."