OpenAI already went through a cycle with the Italian authorities (with presumably the German regulator listening in too). The fix was very simple: give users th

OpenAI already went through a cycle with the Italian authorities (with presumably the German regulator listening in too). The fix was very simple: give users the option not to have their queries used for training. We are still pretty far from an anti-trust case, which is generally harder to navigate. GPDR and other EU regulations on privacy and AI have fairly simple principles: if you do something that is not obvious or necessary (like fraud detection), ask for informed consent. Explain what you are doing, make your steps traceable, allow people to be excluded from non-trivial treatment, to delete the personal information they handed over. It makes some processes a little bit harder to set up, but it’s not really a hard lesson you can only learn by being sued for billions. It’s closer to data engineering 101.