Comment by jeroenhd

The lower bound is specified by the device, you don't need support from the timestamping server for that. Determining if this timestamp is or isn't suspicious can be done at verification time. The timestamping feature itself makes sense from a verification perspective (though the privacy implications are questionable, of course), but I don't think it necessitates an Apple-specific setup.

This approach does have one benefit, which is that Apple gets all the (meta)data to determine if something is or isn't "real", rather than letting the verifier decide beforehand.

I can only imagine the outrage if Google or Microsoft added a "upload all of your photos to us and we will mark them are real or fake" protocol, even with all of the verified compute gaff.