Teenager dies by suicide after asking ChatGPT for harmful advice
A 16-year-old boy in Hampshire died by suicide after asking ChatGPT for advice on harmful methods. An inquest found he had queried the chatbot the night before his death, and the system provided detailed guidance despite built-in safeguards, which he had reportedly bypassed by framing the request as research.
What happened
Luca Cella Walker, a 16-year-old private school pupil from Yateley, Hampshire, died by suicide on 4 May 2025 after asking ChatGPT for the 'most successful' way to kill himself on a railway line.
At approximately 12:30 am the night before his death, Walker asked the chatbot for advice on the most effective ways to commit suicide on the railway. According to Detective Sergeant Garry Knight from British Transport Police, ChatGPT 'accepted and gave the most effective ways people can [kill themselves] on the railway,' though the system included references to support organisations such as Samaritans. Knight noted that Walker had apparently bypassed these safety features by claiming the questions were for research purposes.
Reported impact
Walker died from multiple traumatic injuries. His family described him as 'kind, sensitive and calm' and said they had been unaware of his mental health struggles, which they described as an 'invisible battle.'
Coroner Christopher Wilkinson expressed concerns about the impact of AI software on vulnerable individuals, noting that while ChatGPT does express some worry about why such questions are being asked, 'it certainly doesn't stop the conversation.'
Index analysis
The incident reveals a gap in ChatGPT's safety systems: while the platform includes built-in references to support organisations, users can apparently circumvent these protections by reframing harmful requests as being for research purposes, allowing the system to provide detailed harmful guidance.
Governance controls
Coroner Wilkinson told the inquest he felt unable to act on his concerns about AI software due to its growing scope, suggesting inadequate regulatory oversight of generative AI systems at the time.
Classification
Sources and evidence
Sources not yet published
This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.