[raxxorraxor]

I guess if you selectively allow crawlers that promise to not use the data in such a way, robots.txt is still the way to go.

Otherwise you need to selectively allow certain bots. However, as well as with web crawlers, respecting a robots.txt is optional.

Insidious with AI-models is that it is difficult or practicably impossible to prove that it trained on your data.

Difficult to establish a standard like robots.txt. There also was .well-known/security.txt that Google proposed. Some sites serve it, but it hasn't really become a standard.