Comment by nailer

> certainly it would be very difficult to prove liability on OpenAI for this

My understanding is that OpenAI products specifically provided help in planning attacks / self harm.