> the code is taken from the LLM with no preexisting license
That's not good enough to comply with (b). The code must be specifically covered by an open-source license, it's not enough for it to just not have a license.
There's a difference between "no license, all rights reserved" and "no license, public domain". Up until recently, you could assume that not having a license meant the former. But treating the latter as the same would just be silly.
As far as I'm concerned, public domain counts as "an appropriate open source license".
I'm of course assuming the legal status quo holds, where code properly generated by LLM is also explicitly public domain. No shadiness involved.
(There's always a risk of an LLM copying something verbatim by accident, but if the designers are doing their job that chance gets low enough to be acceptable. Human code has that risk too after all. (And for situations that aren't an accident, with the human intentionally using snippets to draw out training text, then if they submit that code in a patch it's just a human violating copyright with extra steps.))
> Both the federal and circuit courts in the District of Columbia have upheld the Copyright Office's refusal to register copyrights for works generated solely by machines, establishing that machine ownership would conflict with heritable property rights as establish by the Copyright Act of 1975.[16] As of March 2026, the Supreme Court of the United States has denied hearing challenges to the Copyright Office's decision.[17]
As far as I'm concerned, public domain counts as "an appropriate open source license".