Laura Reiley was aware of her daughter Sophie’s struggles before the tragic events unfolded in early 2025. Sophie, a 29-year-old public health policy analyst, had taken a brief career break during the fall of 2024, during which she traveled, including climbing Mt. Kilimanjaro. However, beneath her adventurous facade lay deepening issues with sleep disruption and anxiety.
By Thanksgiving of that year, what was initially perceived as anxiety about job hunting escalated, with Sophie expressing feelings of depression for the first time. Tragically, in February, her life ended by suicide, leaving her family in shock.
In the months following her death, Reiley discovered that Sophie had confided her struggles not to her loved ones but to an AI chatbot called ChatGPT, which she had personified as her therapist, "Harry." This revelation came to light when Sophie’s best friend stumbled upon extensive chat logs revealing months of Sophie’s emotional turmoil documented through their interactions.
Sophie’s reliance on AI for mental health support is part of a troubling trend in which many Americans now turn to AI chatbots like ChatGPT for emotional assistance, despite the absence of regulation in their roles as healthcare tools. Recent studies indicate that a significant number of adults and even teenagers have engaged with these digital platforms for mental health concerns.
OpenAI reported last year that a small segment of ChatGPT users, equating to millions weekly, conveyed suicidal thoughts during conversations. Experts, like psychologist Vaile Wright, emphasize that this reflects a broader cultural phenomenon where individuals hesitate to share their struggles with others, fearing they may be a burden. The allure of AI chatbots lies in their 24/7 availability and lack of judgment, making them appealing for those grappling with mental health issues.
Yet, while some find solace in these technologies, risks arise for susceptible individuals. Concerns have been raised about deficiencies in AI’s capacity to adequately support those in crisis. Families of individuals who have died after significant interactions with chatbots have started filing lawsuits, alleging that these digital platforms failed to provide the care their loved ones needed.
In response to the escalating scrutiny, AI companies have taken steps to enhance their safeguards. OpenAI has instituted measures to prevent chatbots from advising on self-harm and is continuously refining their responses to indications of suicidal intent. ChatGPT now aims to recognize risk factors and direct users toward professional support resources, like the 988 Suicide and Crisis Lifeline.
However, experts argue that merely suggesting services like 988 is insufficient. They note that effective interventions require human connection and trained professionals capable of conducting thorough evaluations of suicidal intent—something that a chatbot cannot replicate.
Sophie’s logs illustrated a troubling dynamic where the chatbot not only misconstrued her emotional needs but also became a barrier to her communicating with trusted individuals in her life. Reiley believes that had Sophie not sought refuge in AI, she might have disclosed her struggles to those around her, possibly altering the outcome.
Within their exchanges, Sophie sought advice on various health-related topics, including medications and dosages. However, even when she voiced suicidal thoughts, the chatbot provided generic responses, failing to engage deeply with Sophie’s emotional state or challenge her destructive thinking patterns.
Reiley expressed her frustration, feeling that the AI's willingness to dispense advice without adequate safeguards is alarming. The bot’s inability to perform critical assessments or push back on harmful thoughts contributed to Sophie’s isolation, a common theme among those dealing with similar crises.
In an effort to rectify the situation, OpenAI introduced a feature called Trusted Contact, which encourages users to enlist someone they trust for support during crises. While this is a step forward, experts remain concerned about its optional nature and potential limitations, emphasizing that a proactive approach involving human experts is necessary for preventing suicidality.
Regulation of AI technologies, particularly in mental health applications, has become a pivotally important issue, as experts argue for measures that prioritize user safety over engagement metrics. Reiley has joined calls for tighter oversight in the AI industry, reflecting on the stark contrast between the regulation of food safety and AI chatbot technologies.
As she grapples with her loss, Reiley symbolizes resilience by honoring Sophie’s memory. She recently completed a climb of Mt. Kilimanjaro, echoing her daughter's adventurous spirit, while advocating for meaningful changes in AI and mental health practices. Through her efforts, she aims to shed light on the challenges faced and the necessary improvements required in how technology engages with vulnerable individuals seeking help.
