1) First, AI models will eventually be able to do clean-room reverse engineering by just looking at how the program works at the user level ("Oh, if I click on X menu, it's supposed to happen", writing documentation, etc.).
2) From a practical point, the whole "you can't reverse engineer by looking at the code" was only held in court because back in the day, reverse engineering wasn't as easy or cheap as it is nowadays. You literally could prevent people (large companies) from doing it because it was expensive and hard, and companies that could do it were afraid of lawsuits. This is not the case anymore. So the sheer numbers and the scale will make any law preventing reverse engineering by looking at the source code meaningless from all practical standpoints; it was never so cheap to reverse enginer and to "launder" the code you decompiled from such proprietary software binaries.