Comment by vostrocity

I think it is a new framing that merits discussion.

Example case is the school shooter in Canada that OpenAI knew about but chose not to warn authorities of (presumably because OpenAI wants to balance safety and privacy).

OpenAI (or any other big tech) has extreme concentration of power and knows more about its users than any government authority.

At what point should OpenAI alert authorities?

I would much rather have "provably beneficial surveillance" than OpenAI having an arbitrary black box policy or for government authority to have direct backdoor to all OpenAI data.