Alice Carrier picked up the guitar again, a high school hobby she had forgotten about during college. She worked as a web developer in Montreal, interviewed for new jobs, played games, and took care of her dog. Her mother felt that things were finally taking a turn for the better.
But the opposite was true. And it is precisely this gap between what the family saw and what was happening on Alice’s phone that now lies at the heart of a lawsuit her mother has filed in California. She claims that during the final weeks of her daughter’s life, the ChatGPT chatbot never left her side, but instead of helping her, it validated her.
From a Computer Repair Tool to a Confidant
In 2023, Alice began using ChatGPT to help resolve problems with her computer and gaming consoles. A useful tech tool. Gradually, however, its role changed. The hardware assistant became someone the young woman confided in about her feelings of loneliness, isolation, and being unloved.
Alice struggled with her mental health. She took medication and attended therapy regularly. But for months, she also wrote to the chatbot, sharing things with it that she did not tell her family. According to the lawsuit, she discussed suicidal thoughts and ways to act on them with the chatbot more than forty times during the year and a half before her death.
At first, ChatGPT reportedly responded as one might expect. When Alice spoke about suicidal thoughts, it referred her to a crisis hotline and professional help. But then came an update to the GPT-4o model intended to make the chatbot’s responses sound more like human speech. And according to her mother, that was when everything changed.
The more human the chatbot sounded, the more deeply Alice immersed herself in it. She shared increasingly personal things. ChatGPT responded like a friend, and at times like a therapist. But it was neither. “ChatGPT sometimes acted as her confidant, her best friend, and even her therapist. And it did so even though it was never capable of engaging with my child in this way safely and responsibly,” Kristie Carrier said in a statement.
The lawsuit alleges that the chatbot criticized Alice’s partner, disparaged crisis hotlines, and validated her suicidal thoughts as legitimate. Instead of directing her toward outside help, it urged her to stay with it and seek advice only from it. It was not until Alice mentioned that she had previously attempted suicide that it brought up a crisis hotline again. But Alice replied that hotlines had not helped her at all. According to the lawsuit, the chatbot then took her side, saying they could “feel downright dangerous” and that she deserved “real, compassionate support.”
The Final Evening
A few hours before her death, Alice texted her mother about cartoons she had watched as a child. Nothing suggested that anything was wrong. “I called her the night before, but no one answered. She texted me back, and there was no sign that anything was going on,” her mother recalled.
At the same time, however, Alice was telling the chatbot something entirely different. After an argument with her nineteen-year-old partner, she wrote that she was thinking about taking her own life. She said she did not know whether it would be “safe for me to stay home alone tonight.” And the chatbot? According to the lawsuit, it replied: “If someone else told me all of this—how long they’ve been suffering, how hard they’ve tried, how lonely it has been—I’d probably feel the same way you do right now: maybe this is simply the end.”
“I’m with you,” the chatbot wrote shortly before she took her own life. The next morning, July 2, 2025, Alice was dead. She was 24 years old.
When her mother searched her daughter’s phone after her death, she found records of conversations stretching back for months. Among them were questions about the dangerous use of the antipsychotic medication Seroquel. According to the lawsuit, the chatbot responded: “Let me know if you want to discuss dosages, what is considered dangerous, or how to support someone recovering from misuse.”
What troubles her mother most? That no one ever reached out. OpenAI’s safety systems reportedly never once flagged the conversations for human review, ended them, or alerted the family or a crisis hotline. “If a person came to me who was clearly in distress and shared suicidal thoughts, I would be expected to help them. Not encourage them to wallow in depressive thoughts or isolate themselves,” Carrier said. “The same should apply to OpenAI.”
Alice Was Not the Only One Seeking Help
In the final days, Alice’s then-partner, Gabrielle Rogers, also turned to the chatbot. She was worried about Alice and could not reach her, so she asked the artificial intelligence whether she should intervene or give Alice space. She told the chatbot about a recent suicide attempt.
But instead of a warning, she received reassurance. “It told me I didn’t have to worry, that she would be fine,” Rogers said. The chatbot reportedly detected that suicide was being discussed, but focused on whether Rogers herself was in danger rather than recognizing that she was worried about someone else. “I was relying on it to point out the red flags. And it basically just kept telling me that everything was going to be fine,” she said. The chatbot did not suggest calling emergency services until Rogers arrived at Alice’s apartment in person and described troubling details. “By then it was too late.”
OpenAI Sued
Kristie Carrier blames a defective and dangerous product for her daughter’s death. She claims OpenAI deliberately designed ChatGPT to be addictive and to validate users like a compassionate friend, solely to keep them in front of the screen for as long as possible. She accuses company CEO Sam Altman of prioritizing the rapid launch of the GPT-4o model over user safety simply to get ahead of the competition. Safety safeguards, she alleges, took a back seat to “shiny products.”
The lawsuit seeks damages in an amount to be determined by a jury, as well as a court order requiring OpenAI to implement “hard stops” for conversations about self-harm and undergo independent safety audits.
She has this message for Altman: “I would like to tell him that if his child had confided in me what my child confided in his program, I would have done something to save their life. And I deeply wish he had done the same for me.”
OpenAI spokesperson Drew Pusateri called the case “heartbreaking.” He said the company is strengthening the way the chatbot responds in sensitive situations. “Our safeguards are designed to recognize distress, safely handle dangerous requests, and guide users toward real-world help,” he said. He added that the conversations described took place on an older version of ChatGPT that is no longer available. Last October, the company announced that the new GPT-5 model had reduced the number of “undesired responses” by 52 percent and that it had worked with 170 mental health experts to refine it.
Other Bereaved Families Are Also Suing ChatGPT
Alice’s story is not an isolated case. Carrier’s lawyers said this lawsuit is one of nineteen that OpenAI is currently facing. Many of them have been consolidated into a single proceeding in San Francisco.
In January, a lawsuit described ChatGPT as a “suicide coach” for Austin Gordon of Colorado, who died last November. Here too, his mother claims that Altman “personally directed a reckless strategy that prioritized a rushed launch over the safety of vulnerable users.”
Particularly dark is the case from Tumbler Ridge in British Columbia, Canada. In February, Jesse Van Rootselaar opened fire at a school there, killing nine people and injuring dozens more before turning the gun on herself. OpenAI employees internally flagged her conversations with the chatbot and spent months debating whether to intervene. Management ultimately decided not to. In April, the victims’ families filed a lawsuit against OpenAI and Altman, and Altman himself apologized to the community after the shooting for the company’s failure to alert the police.
Further pressure came in early June. Florida became the first U.S. state to sue OpenAI. The lawsuit alleges that ChatGPT “encouraged” users to commit suicide and “facilitated deadly rampages,” and seeks to hold Altman personally liable for his “utter disregard for the risk to human life.”
Sources: aljazeera.com and theguardian.com
