To me it seems like these concerns should have been brought up when implementing the AI policy.
In fairness, @AaronBallman did bring up his concerns early in the original AI policy proposal.
As I have mentioned upthread, I believe it is extremely unlikely for these patches to contain large swathes of code duplicated exactly from other codebases given how specific the patches are to LLVM.
@chandlerc mentioned during the AI policy roundtable at the 2025 LLVM dev meeting that the copyright status of AI generated code is well-settled, but only in a US context. See my notes at AI policy roundtable meeting notes