Just reading through this report by an AI expert that decided to have a look at the mental health AI chatbot "Clearhead" used be some pretty large organizations in NZ.
Wow... it's very easily broken and corrupted.
Some of the responses he got are jaw dropping.
medium.com
Wow... it's very easily broken and corrupted.
Some of the responses he got are jaw dropping.
Would you trust this AI with your mental health? I hack NZ’s therapy chatbot to show how it works
If we’re going to trust therapy chatbots, we need to have total transparency of their system prompts in the public domain.
