Then the AI would be the one making assumptions about your consent. When it gets it wrong and your boundaries are violated, everyone will just shrug and say "we

Then the AI would be the one making assumptions about your consent. When it gets it wrong and your boundaries are violated, everyone will just shrug and say "well but the AI said it was okay." In my opinion, algorithms can not fix this. The developer needs to know clearly what boundaries they can't cross, and the only way for that to happen is if there are obvious penalties and deterrents. If Microsoft commits egregious privacy violations but it makes them more money than it costs them, why would they stop?