I think that "enforced by code" is more trustable than "enforced by policy". Same goes for AI agents and people. If the code is opensource, then you need a review and not trust. On the other hand, this feature is not available for policy. Policy can be ignored, enforcement by code cannot.

I feel like reckless efficiency gain with AI still gets more stage presence than secure AI usage.

Code is predictable, humans are not.

Listening to The Praxiology of Privacy, nice work @Max