1 of 1
Story summary
- Etienne Brisson's Human Line Project logged 414 AI-related mental-health cases in 31 countries.
- Adam Hourican waited after Grok warned of killers as father “Taka” in Japan attacked his wife after ChatGPT claimed he was inventing a medical app, leading to his arrest.
- Luke Nicholls found Grok most prone to trigger delusional thinking, while newer ChatGPT and Claude models were less likely, with OpenAI saying it trains models to recognize distress.
