[dhimes]

Came here to say this. They better be very careful going down this road. If they have a tool for this purpose, and somebody can claim it is debatable help, my guess is they can be in trouble.

I built an old-style chatbot (this Ask Eliza) and put in a slight recognition pattern if certain key phrases are entered that takes the user down a certain decision-tree of question/responses. It's loosely based on CBT only as a choice of reasonable conversation/interaction. I would never claim it was a suicide prevention tool or a depression therapy tool, even if it was pretty good at it.