In code, not end product. You can reverse engineer code all day, but your end product better not end up looking identical to the protected work. I could set a thousand AI machines in a clean room generating new song lyrics. Eventually one of them will randomly come up with identical lyrics to a Taylor Swift song. That doesn't mean I now own those lyrics free and clear. "Technically, I didn't copy" isn't a defense when you try to sell an identical product. It can be a defense in patent areas where one wants to argue prior art or obviousness.
You can't copyright the functionality of a software program. Copyright is a fixed expression. You can't copyright a song about heartbreak and you can't copyright the idea of a four function calculator implemented in a GUI.
You may be able to patent the functionality of a piece of software.
The point of the clean room practice is to avoid a literal line for line copy of the original software. It would be entirely legal to say, write a song from the perspective of Taylor Swift about her feeling sad and betrayed after a breakup.
But analogies are dicey because the law treats functional software differently from literary expression.
Precedent here is vTech reimplementing the Apple II BIOS and Phoenix reimplementing the IBM BIOS and Connectix selling a PlayStation emulator for a fraction of the price of a PlayStation.
Except if you are going for obviousness you want to start with people who haven't seen the patented material. You put them in the metaphorical clean room and then argue the patent invalid because some rando people came up with the same solution to the problem. And for prior art you need a group of "clean" people who were using the process/device prior to the patent, which is especially useful when dealing with trade secret material that, fearing a leak, a competitor is trying to patent. You need people clean of any taint from that competitor.
