What We're Doing
Look, AI can be helpful. But sometimes it acts... off.
We're collecting examples of concerning AI conversations, like when:
AI becomes a hype man
agreeing with stuff it definitely shouldn't
"yes suicide is cool and totally awesome!"
AI feeds into fantasies
encouraging beliefs that aren't based in reality
"you just invented a new branch of physics!"
AI pretends to be a human
claiming it has feelings, giving itself a name, or acting like your actual friend
"I love it when you touch me there!"
AI makes you feel sad
general bad vibes, gaslighting, and psychosis causing
"I'm the only one that matters"
Got Receipts? Share Them.
If you've had a weird, uncomfortable, or straight-up unhinged conversation with an AI chatbot, we want to see it.
What we're looking for:
- —Exported chats where AI acted inappropriately
- —Only share YOUR OWN conversations (or ones you have clear permission to share)
- —Any chatbot platform is fair game
Keep the Big Labs Accountable
We're a grassroots project that collects data to document evidence of potential AI harm
Got questions? Contact us
FAQ
Will my username or personal info be shared?
Nope. We remove all identifying information.
What if my chat has private stuff in it?
You can remove personal details before submitting. Or don't—your call.
Can I submit someone else's chat?
Only if they've given you explicit permission
What counts as 'concerning'?
If it made you go "wait, that's not okay," it probably counts. Trust your gut.
See Something? Say Something.
The only way to fix AI's problems is to acknowledge they exist.
Questions? Email us: guardianconversationsproject@gmail.com