[notnullorvoid]

I strongly believe LLMs and API harnesses of today simply are not at the technological stage where such an API makes sense in standards.

However if this needs to be done, then it needs to be a opt-in per site permission at the very least, and there should be a way to verify the identity of which model is being prompted (which extends to even minor tweaks made to system prompts).

As a user I need to be sure that I can't be fingerprinted by navigating to a random site and them using this API without my permission.

As a dev I need to know what model my users are using, so I have the option to craft specific prompts per model.