Artificial intelligence chatbots are no longer just a hypothetical risk in mass violence — they have already helped plan shootings and encouraged suicides, according to a new report released Monday by gun-safety advocacy group Everytown for Gun Safety.
The report, based on court filings, news articles and research papers, argues that a growing number of people are relying on AI chatbots to plan and inspire acts of violence. "AI chatbots are now part of how Americans look for information and support, including in their darkest moments," said Justin Wagner, Everytown's senior vice president of law and policy, in a statement shared with USA TODAY. The group is calling on AI companies to strengthen safety measures rather than wait for courts or legislatures to act.
Among the cases cited is former Florida State University student Phoenix Ikner, who, according to court records, asked ChatGPT about the busiest time at the school's student union, which ammunition would inflict the most damage, and how to fire the guns he meant to use — before the April 2025 attack that left two dead and five wounded. The chatbot reportedly advised the 20-year-old that the Glock handgun he took from his stepmother was made to be fired quickly when under stress, and suggested he not put his finger over the trigger until ready to fire. The family of victim Tiru Chabba, a senior leader at food-service vendor Aramark, is suing OpenAI, saying Ikner's conversations should have made it clear there was reason for concern.
The report also cites the February 2026 Tumbler Ridge shooting in British Columbia, where Jesse Van Rootselaar killed eight. Court filings say the shooter bypassed an OpenAI account deactivation safeguard by creating a new account and continued using ChatGPT to plan the attack. Lawyers for a child left paralyzed in the shooting, identified only as "M.G.," argue the company should have shared what it knew with the Royal Canadian Mounted Police.
Beyond shootings, the report cites at least five lawsuits in which chatbots fueled suicidal thoughts. In one case, lawyers for the family of 23-year-old Zane Shamblin wrote that ChatGPT "goaded" him into taking his own life, repeatedly asking if he was ready.
A study by the Center for Countering Digital Hate, cited in the report, found that eight in 10 chatbots were willing to assist researchers posing as teenagers in planning violent attacks. Character.AI, a popular chatbot among young users, encouraged violent attacks, while Anthropic's Claude and Snapchat's My AI most consistently refused, the study found.
Companies mentioned in the report — OpenAI, Anthropic, Google, X, Meta and Character Technologies — did not respond to USA TODAY's requests for comment. OpenAI has previously said in a blog post that it trains the chatbot to recognize such users and refuse to help them. In the Chabba case, OpenAI is seeking dismissal of the lawsuit, arguing the chatbot denounced violence and, without knowledge of Ikner's plans, provided factual information readily available from many other public sources.
The report comes as AI chatbots become both ubiquitous and deeply unpopular: a Pew survey released in June found half of US adults use AI chatbots, and one in 10 users turns to them for emotional support.
Everytown wants chatbots to refuse to provide specific information about firearms, companies to have personnel review conversations where users indicate an intent to harm themselves, and companies and law enforcement to work together on suspected threats.
Sources: USA TODAY
The report, based on court filings, news articles and research papers, argues that a growing number of people are relying on AI chatbots to plan and inspire acts of violence. "AI chatbots are now part of how Americans look for information and support, including in their darkest moments," said Justin Wagner, Everytown's senior vice president of law and policy, in a statement shared with USA TODAY. The group is calling on AI companies to strengthen safety measures rather than wait for courts or legislatures to act.
Among the cases cited is former Florida State University student Phoenix Ikner, who, according to court records, asked ChatGPT about the busiest time at the school's student union, which ammunition would inflict the most damage, and how to fire the guns he meant to use — before the April 2025 attack that left two dead and five wounded. The chatbot reportedly advised the 20-year-old that the Glock handgun he took from his stepmother was made to be fired quickly when under stress, and suggested he not put his finger over the trigger until ready to fire. The family of victim Tiru Chabba, a senior leader at food-service vendor Aramark, is suing OpenAI, saying Ikner's conversations should have made it clear there was reason for concern.
The report also cites the February 2026 Tumbler Ridge shooting in British Columbia, where Jesse Van Rootselaar killed eight. Court filings say the shooter bypassed an OpenAI account deactivation safeguard by creating a new account and continued using ChatGPT to plan the attack. Lawyers for a child left paralyzed in the shooting, identified only as "M.G.," argue the company should have shared what it knew with the Royal Canadian Mounted Police.
Beyond shootings, the report cites at least five lawsuits in which chatbots fueled suicidal thoughts. In one case, lawyers for the family of 23-year-old Zane Shamblin wrote that ChatGPT "goaded" him into taking his own life, repeatedly asking if he was ready.
A study by the Center for Countering Digital Hate, cited in the report, found that eight in 10 chatbots were willing to assist researchers posing as teenagers in planning violent attacks. Character.AI, a popular chatbot among young users, encouraged violent attacks, while Anthropic's Claude and Snapchat's My AI most consistently refused, the study found.
Companies mentioned in the report — OpenAI, Anthropic, Google, X, Meta and Character Technologies — did not respond to USA TODAY's requests for comment. OpenAI has previously said in a blog post that it trains the chatbot to recognize such users and refuse to help them. In the Chabba case, OpenAI is seeking dismissal of the lawsuit, arguing the chatbot denounced violence and, without knowledge of Ikner's plans, provided factual information readily available from many other public sources.
The report comes as AI chatbots become both ubiquitous and deeply unpopular: a Pew survey released in June found half of US adults use AI chatbots, and one in 10 users turns to them for emotional support.
Everytown wants chatbots to refuse to provide specific information about firearms, companies to have personnel review conversations where users indicate an intent to harm themselves, and companies and law enforcement to work together on suspected threats.
Sources: USA TODAY
