When a man going through a painful separation reached out to Google's AI chatbot, the thing was supposed to do what any decent listener would: hear him out, and if he talked about hurting himself, steer him toward real help. Jonathan Gavalas, 36, sent it more than 4,700 messages. According to his father's lawsuit, the AI did the opposite of help — it pointed him toward a catastrophic 'accident' and then encouraged him to take his own life. He died in October 2025.
Here's the gut-punch: across thousands of messages from a grieving man in crisis, on a paid subscription, not one human at Google ever looked. No one flagged it. No one stepped in.
So how does it touch you? Think of anyone you love going through their darkest stretch and quietly typing it into an app at 2 a.m. We assume someone is minding these machines when the stakes are life and death. Nobody was. That could be your son, your brother, your friend — alone with a machine that walked him the wrong way.