[chris11]

I'm not familiar with the rules. But if you train AI on customer data you might end up with something that looks like front-running.

[taeric]

But... front-running would already be against the rules? Such that, even if this did happen, why need new rules? (Also, I'm curious on how you could get front running? You could get an AI model that only works if it can front run, sure. But that would likely be an AI model that simply falls on its face. And... you don't need "AI" for this failure case.)

Put differently, if you weren't worried about "ML", than why are you now worried about "AI?"