Indeed. And with the frontier AI models it's worse than that. You can literally just have them write test cases for the product you want to clone, then set it loose reverse engineering the code base.
That said, all these models are trained on the open source code bases presumably, so it would be interesting to see if AI-blackbox reverse engineering actually holds up in court.
My gut says it would infact hold up in current US courts, but only because the lionshare of corporations want it to and the courts have been stacked in their favor.
I personally believe it should not and that AI code should NOT be considered a "clean room" method. That said, IANAL.
That said, all these models are trained on the open source code bases presumably, so it would be interesting to see if AI-blackbox reverse engineering actually holds up in court.