While I feel sorry for this engineer's ordeal, if you insist on seeking health advice from a "low-latency, fast model" like 4o, you're indeed being irresponsible with your health. GPT-4 already existed when 4o was developed; if this news was from before last year, GPT-4.5 and even GPT-5 were already available. If you had consulted these models, this tragedy wouldn't have occurred. 4o is more like emotional companionship. (For a long time, GPT's voice mode was based on 4o, emphasizing rambling but instant responses.)