ChatGPT blocked 50,000 public postings from Google but overlooked 100,000 more. Sampling these accidentally public conversations, I found a digital confession booth: mental health struggles, personal secrets, criminal admissions.

People treating ChatGPT as therapist—sharing details they'd never tell humans. Tax schemes, betrayals, crises, illegal activities.

It's like finding 100,000 photos on the street, discovering some show crimes. These weren't public statements—digital whispers that accidentally became shouts.

Users clicked "Share" but most had no clue what that meant. Interface confusion turned private conversations into public confessions. When 100,000 people make the same "mistake," it's system failure, not user stupidity.

We protect free speech and privacy, even when speech describes illegal acts. Someone confessing tax evasion isn't automatically prosecutable—confession alone often isn't evidence.

But what happens when private confessions become accidentally public? In authoritarian states, this would be surveillance goldmine. What democracies protect becomes persecution grounds.

ChatGPT can't pre-screen every conversation without destroying privacy. Real-time monitoring would turn every AI chat into police informant.

But once content becomes accidentally public, they face pressure to act. Should they report crimes? Delete everything?

This reveals people treat AI chatbots like digital therapists. Are we building systems that make privacy accidents inevitable?