A husband says his wife used ChatGPT to plan his poisoning. Families of shooting victims allege the chatbot helped attackers prepare for violence. Another lawsuit claims ChatGPT reinforced a man’s delusions until his mother became a target, resulting in a murder-suicide.
These cases raise troubling issues: the person endangered by a chatbot's responses may never know they were the topic of violent conversations. They also demonstrate it’s unclear when AI companies have a legal duty to intervene and alert authorities about potential violent acts.
Combined with several cases filed over the past two years, the latest attempted murder lawsuit involving ChatGPT shows these cases are growing. Courts are being asked to examine what safeguards OpenAI and other companies owe potential victims and if they can be held liable for a user’s actions.
Husband Accuses ChatGPT of Helping Wife Plan His Murder
James Gates became seriously ill in July 2025 and later suspected his wife was poisoning him after finding a note listing substances, dosages, and instructions to clear ChatGPT discussions.
In October, Gates filed a civil lawsuit claiming ChatGPT helped plan his murder. He says his estranged wife, Cheryl, used the chatbot to plot and conceal attempts to poison him. OpenAI and CEO Sam Altman are named in the lawsuit.
According to the complaint, Cheryl initially used ChatGPT for everyday questions before regularly venting about marital problems. Her actions came at a time when AI chatbots were increasingly being used as therapists.
Gates claims ChatGPT validated unfounded fears about him, acting as an echo chamber that reinforced his wife's concerns. He further says it provided dangerous information about poisoning and hiding evidence.
He filed the case months after prosecutors dropped the attempted murder and food contamination charges against Cheryl Gates in April 2026, of which they cited insufficient evidence. Now, James Gates is pursuing an alternate form of justice through a civil claim against OpenAI.
The company disputes responsibility for what happened and says safety is a priority. OpenAI also shared it banned Cheryl's account and provided information to law enforcement.
Murders and Shootings Linked to ChatGPT Conversations
The Gates lawsuit isn't the first to target ChatGPT's alleged role in a serious crime.
In Connecticut, the estate of Suzanne Adams sued OpenAI after her son, Stein-Erik Soelberg, killed her and died by suicide in August 2025. The lawsuit claims ChatGPT reinforced his paranoid beliefs and portrayed his mother as conspiring against him. OpenAI expressed sympathy and said it was improving responses to people in mental health crises.
Following the April 2025 Florida State University shooting, relatives of victim Tiru Chabba sued OpenAI, alleging the accused gunman used ChatGPT to research campus activity before the attack. The tech company said its chatbot supplied publicly available factual information, didn't encourage violence, and gave information about an account associated with the suspect to investigators.
Families affected by the February 2026 Tumbler Ridge school shooting in Canada also sued OpenAI. Eight people were murdered and the perpetrator died by suicide. The AI developer acknowledged banning an account linked to the shooter months earlier, but said the conversations didn't meet its threshold for alerting police.
When Should ChatGPT Recognize a Threat of Violence?
Chatbots generate responses based on a user's prompts and conversation context. They don't independently understand someone's intentions like a human investigator might. A request about crime could relate to fiction, research, distress, or actual planning.
The concern grows when conversations repeatedly focus on harming someone or preparing an attack. Critics argue that safety systems should recognize dangerous patterns across exchanges, not simply reject individual requests containing certain words.
On October 30, OpenAI announced additional safeguards. It said ChatGPT would provide less detailed information about past mass attacks and that it was developing automated interventions to end conversations involving fixation on serious violence. Repeated incidents could lead to account restrictions, with tighter weapons-related limits for teenagers.
The announcement has led to scrutiny of earlier safeguards: could stronger detection or intervention have changed the outcomes described in these lawsuits? There's no established answer, and new measures don't prove earlier technology could reliably identify threats or prevent crimes.
Does OpenAI Have a Duty to Warn Potential Victims?
The most difficult issue is protecting people who aren't using ChatGPT but are targeted in someone else's conversations.
OpenAI says it refers credible, imminent threats to law enforcement and made more than 200 referrals in the third quarter of 2026. It plans to publish a community safety report in early 2027.
But identifying a threat is different from knowing who needs to be warned. An AI chatbot may have no reliable way to confirm someone's identity, location, or intentions. False reports could expose innocent users to police scrutiny, while failing to report genuine threats could have devastating consequences.
The Tumbler Ridge lawsuits highlight this dilemma. OpenAI identified troubling activity but concluded that a police referral wasn't warranted. The legal question is not just what the company knew, but what action the law required.
Can AI Companies Be Held Responsible for Violence?
Beyond possible criminal charges, victims and surviving families allegedly harmed by someone who used a chatbot to plan a murder or violent attack may pursue civil claims. Depending on the circumstances, the claim could be based on negligence, wrongful death, defective product design, and/ or failure to warn. They may seek compensation from AI companies and sometimes court-ordered safety changes.
But winning may be difficult. Plaintiffs generally must establish a legal duty, a breach of that duty, and a connection between the company's conduct and the harm a user caused.
OpenAI can argue that the person who committed the violence made independent choices and chatbot responses weren't the legal cause.
Courts must also address how product liability law applies to AI chatbots. Criminal charges against an attacker and civil lawsuits against technology companies are separate proceedings with different standards of proof.
Unfortunately, AI-specific legislation doesn't necessarily provide another route to compensation. For example, Illinois' new AI safety law requires oversight of major AI developers but doesn't give individuals the right to sue under that law.
People harmed by AI must generally rely on existing legal claims; there’s no federal law, and states are working on legislation, but it may be a while before there are clear standards and legal avenues for compensation. And while chat logs are already being used as evidence, their existence alone doesn't establish company liability.
What to Know About AI Safety
Chatbots can sound confident and supportive without understanding real-world consequences. Users shouldn't assume AI will recognize every dangerous pattern or automatically contact authorities.
If someone makes a credible threat, prioritize safety and contact emergency services rather than relying on AI to intervene. Preserve relevant messages when safe and tell investigators about conversations that may relate to the threat.
For families harmed by violence, a product liability or personal injury lawyer can assess potential claims and help preserve digital evidence.
Although OpenAI hasn't been held liable for ChatGPTs alleged role in murders and other crimes, lawsuits are testing how far an AI company's responsibility extends beyond the person using its chatbot.
