The DSA mandates algorithmic transparency. In theory they could demand that the adds served to any user be explained in detail. Just create an ensemble of user

The DSA mandates algorithmic transparency. In theory they could demand that the adds served to any user be explained in detail. Just create an ensemble of users, make them differ only on things meta is not supposed to use, demand in detail explanations for the different behaviour of the system. If the sample size is large enough you can reject the null hypothesis that certain information was not used.