I want to add that AI-generated code isn't FOSS because it's not a _source_ code at first place. This is a generated code. All other reasons, in my opinion, are secondary and may change over time.
Publishing generated code is not unique for AI. For example, it's a common practice to generate a parser using YACC from source grammar files, and then publishing both. The main thing here is that no one claims that this code is a "source code", no one saying that it has been written by someone. It is perfectly fine to put a comment directly in the file's text that it has been generated by YACC.
If people would be more accurate on this matter, and wouldn't publish generated code alongside with OSS projects, a notable part of the problems we discuss today wouldn't exist.
Publishing generated code is not unique for AI. For example, it's a common practice to generate a parser using YACC from source grammar files, and then publishing both. The main thing here is that no one claims that this code is a "source code", no one saying that it has been written by someone. It is perfectly fine to put a comment directly in the file's text that it has been generated by YACC.
If people would be more accurate on this matter, and wouldn't publish generated code alongside with OSS projects, a notable part of the problems we discuss today wouldn't exist.