IANAL but if your only source for your LLM is that code, I would assume the code it produces would be at high risk of being counterfeit.
I would guess clean room would still require having someone reading the LLM-decompiled code, write a spec, and have someone else write the code.
But this is definitely a good question, especially given the recent court verdicts. If you can launder open source licensed code, why not proprietary binaries? Although I don't think the situation is the same. I wouldn't expect how you decompile a code matters.
I would guess clean room would still require having someone reading the LLM-decompiled code, write a spec, and have someone else write the code.
But this is definitely a good question, especially given the recent court verdicts. If you can launder open source licensed code, why not proprietary binaries? Although I don't think the situation is the same. I wouldn't expect how you decompile a code matters.