AI Companies Face Legal and Ethical Challenges Over Chatbot Violence Reporting

The deadly Tumbler Ridge shooting in Canada has raised questions about when AI companies should report threats to police and the risks to individual privacy. On February 10, 2026, 18-year-old Jesse Van Rootselaar shot his mother and half-brother at home before going to a school in Tumbler Ridge, British Columbia. Van Rootselaar killed another six people, including five students, at the school before taking his own life. The attack was Canada's deadliest school shooting in almost four decades. According to reports, Van Rootselaar consulted OpenAI's ChatGPT for months about perpetrating mass murder, and ChatGPT advised Van Rootselaar about a shotgun's suitability for the attack. US authorities on October 14, earlier this week, charged a Washington State man with allegedly conspiring with Van Rootselaar to plan the attack.

The British Columbia government filed a civil lawsuit against OpenAI and CEO Sam Altman in September, alleging knowledge of the conversations and failure to prevent the incident. Families of the victims filed separate lawsuits. OpenAI's leadership publicly apologised for failing to notify authorities. OpenAI has not responded to requests for comment on the lawsuits.

In August, Anthropic reported a user to San Francisco police after discovering conversations allegedly threatening Anthropic CEO Dario Amodei. The San Francisco Standard, citing a police report, said the user claimed to possess an AR-15 and suggested Amodei was a target. However, the man told the newspaper that the comments were jokes and that the man had no intention of committing violence. The Standard said the man was neither charged nor detained.

Fazl Barez, a senior AI safety and governance researcher at the University of Oxford, said current AI models cannot reliably distinguish hypothetical interactions from discussions likely to lead to harm. Barez said an excessively low reporting threshold could wrongly flag fiction writers exploring murder plots or graduate students conducting legitimate research on dangerous pathogens. Barez said credible threat assessment may require examining multiple conversations rather than isolated statements. Barez said distinguishing someone seeking support for intrusive violent thoughts from someone gradually revealing attack plans requires consideration of context, intent and behavioural patterns.

Chinmayi Sharma, a Fordham University law professor, compared the debate with mental health professionals' duty to warn identifiable third parties about credible threats. However, Sharma said therapists have professional training, knowledge of patients and legally enforceable standards that AI companies do not necessarily have. Sharma said AI companies lack a concerted effort to hire mental health professionals, sociologists and other experts who could identify threat indicators. Sharma warned that models trained to identify dangerous speech would almost always reflect biases, including disproportionately flagging certain minorities' speech. Sharma said machines could reproduce unequal outcomes at scale even without malicious intent.

Millions of people use chatbots to discuss relationships, emotional distress, mental health concerns and deeply personal experiences. Jenin Younes, a US civil liberties lawyer and president and legal director of the American-Arab Anti-Discrimination Committee, said most users assume nobody else will view those conversations. Younes said requiring companies to disclose threats is extraordinarily difficult, especially when companies do not already screen chats for threats. Younes argued that forcing companies to create surveillance systems raises serious civil liberties concerns, including US constitutional protections against unreasonable searches. However, Younes acknowledged that AI provides access to information that may facilitate mass-casualty events, a concern that cannot easily be dismissed. Younes warned that fear of police disclosure over disturbing thoughts or violent fantasies could discourage people from using chatbots.

Legal experts say AI companies' legal obligations to prevent violence are less clear than the moral pressure companies face. Bryan Choi, a University of Colorado Boulder law professor, said US law recognises duties to warn in some circumstances involving imminent threats or special relationships. Choi said current case law appears unlikely to impose such duties on OpenAI because chatbots are not licensed therapists and OpenAI has no special relationship with victims. Choi said negligence and product liability laws could potentially apply to AI companies, but proving responsibility would pose significant challenges.

Sharma suggested that independent institutions could establish reporting standards rather than leaving decisions entirely to companies or requiring immediate police disclosure. Sharma said those institutions could develop threat-assessment procedures and explore sharing only relevant information while protecting users' identities and other personal details. Younes argued that users should at least receive clear information about monitoring and the circumstances that could trigger disclosure to law enforcement. Younes said the issue would be debated over coming years, no solution was currently clear, and future events would likely influence policy.