If you have a broken system, and you allow people in power to fix it in targeted ways that only affect them, then it will never get fixed for ordinary people.

Your point might hold true if the people being exempted from surveillance weren't also the people deciding who can be exempted from surveillance. But since they are, it is counterproductive to the privacy cause to allow Congress to give itself these kinds of special rules.

Force everyone in society to play by the same rules, and everyone in society will get the benefits. Allow Congress to get into the habit of exempting themselves from social systems or problems, and those problems will continue to exist.

In other words, when setting up incentives you have to think pragmatically about the entire system in the long term; you can't always just focus on short-term goals to the exclusion of everything else. Local optimums[0] aren't just a problem in AI, they're also a problem in real life. Because of the long term consequences, we're not interested in forming privacy castes in scenarios where doing so will make it harder for us to improve the system in the future.

[0]: https://en.wikipedia.org/wiki/Local_optimum