@tuban_muzuru The guarantee is that the parts of the codebase that aren't written by humans are not called out. You can read into the rest however you like.

@jamie

.... how can you distinguish between 'em?

@tuban_muzuru You came into my mentions with guns blazing, so you'll have to forgive me if I don't consider that to be an honest question.

@jamie

Stop whining. You and about seventy zillion terrified sheep running around here bleating about the Terrible AI monster under the bed.

FWIW I'm not a lawyer and I'm not recommending that you do this. Even if companies have no legal standing on copyright, their legal team will try it. It *will* cost you money.

But man, oh man, I'm gonna have popcorn ready for when someone inevitably pulls this move.

If you use AI-generated code, you currently cannot claim copyright on it in the US. If you fail to disclose/disclaim exactly which parts were not written by a human, you forfeit your copyright claim on *the entire codebase*.

This means copyright notices and even licenses folks are putting on their vibe-coded GitHub repos are unenforceable. The AI-generated code, and possibly the whole project, becomes public domain.

Source: https://www.congress.gov/crs_external_products/LSB/PDF/LSB10922/LSB10922.8.pdf

If you use AI-generated code, you currently cannot claim copyright on it in the US. If you fail to disclose/disclaim exactly which parts were not written by a human, you forfeit your copyright claim on *the entire codebase*.

This means copyright notices and even licenses folks are putting on their vibe-coded GitHub repos are unenforceable. The AI-generated code, and possibly the whole project, becomes public domain.

Source: https://www.congress.gov/crs_external_products/LSB/PDF/LSB10922/LSB10922.8.pdf

If you use AI-generated code, you currently cannot claim copyright on it in the US. If you fail to disclose/disclaim exactly which parts were not written by a human, you forfeit your copyright claim on *the entire codebase*.

This means copyright notices and even licenses folks are putting on their vibe-coded GitHub repos are unenforceable. The AI-generated code, and possibly the whole project, becomes public domain.

Source: https://www.congress.gov/crs_external_products/LSB/PDF/LSB10922/LSB10922.8.pdf

@jamie the corporations own the Supreme Court of the US who will cheerfully make up new law out of whatever Clarence Thomas shat that morning.

@jamie
That doesn't sound quite right. The code "AI" generates are stolen from other people, presumably they still have their copyright.

Also, even if we pretend that it is actually generated by a machine, so is the output of a compiler. Is Microsoft Office in the public domain, because the executable was generated by a compiler?

@jamie Microslop, allegedly, vibe coded up to 30% of the Windows 11 codebase using genAI. Theoretically anyone working for Microslop with access to the codebase can upload the whole codebase without recourse?

If you use AI-generated code, you currently cannot claim copyright on it in the US. If you fail to disclose/disclaim exactly which parts were not written by a human, you forfeit your copyright claim on *the entire codebase*.

This means copyright notices and even licenses folks are putting on their vibe-coded GitHub repos are unenforceable. The AI-generated code, and possibly the whole project, becomes public domain.

Source: https://www.congress.gov/crs_external_products/LSB/PDF/LSB10922/LSB10922.8.pdf

@leeloo Last I heard, the holders of the copyrights on the material that the LLMs are trained on are being told to get fucked.

The class action lawsuit that Anthropic lost was decided not because they trained their models on stolen copyrighted material, but because they stored copies of that material to keep training their models on. My understanding is that it was the storage specifically that violated copyright and that, if they'd deleted that data they'd have been legally clear.

@jamie @aeva I suspect that courts would not be favorable to this reading, and would buy the (bullshit, IMO) argument that sufficient human interaction with the code "heals" the copyrightability of the result, and more importantly that they would not press the applicant to show much work when it comes to "sufficient" (that is, I suspect many judges would accept "I edited the code at all" as meeting the sufficiency criterion)

but we're only going to find out if and when it's tested. The Copyright Office is doing the best they can do and making it clear that they won't let "AI" waste their time with copyright registrations (which are not required to legally protect a work, they're just paperwork really)

@SnoopJ @jamie i think the point is this is a possible interpretation that has precedence already

If you use AI-generated code, you currently cannot claim copyright on it in the US. If you fail to disclose/disclaim exactly which parts were not written by a human, you forfeit your copyright claim on *the entire codebase*.

This means copyright notices and even licenses folks are putting on their vibe-coded GitHub repos are unenforceable. The AI-generated code, and possibly the whole project, becomes public domain.

Source: https://www.congress.gov/crs_external_products/LSB/PDF/LSB10922/LSB10922.8.pdf

@SnoopJ @jamie @aeva That was my read as well. IANAL, but my lay understanding was that even if the courts eventually don't act favorably towards an argument, that it exists and has precedent is enough to create legal risk?

If you use AI-generated code, you currently cannot claim copyright on it in the US. If you fail to disclose/disclaim exactly which parts were not written by a human, you forfeit your copyright claim on *the entire codebase*.

This means copyright notices and even licenses folks are putting on their vibe-coded GitHub repos are unenforceable. The AI-generated code, and possibly the whole project, becomes public domain.

Source: https://www.congress.gov/crs_external_products/LSB/PDF/LSB10922/LSB10922.8.pdf

@xgranade @jamie @aeva I think it's a much stronger case for the example rejected registrations that they show on the next page, which are exclusively about copyrightability of images.

It's largely legally untested AFAICT but based on how eagerly US courts have swallowed up the fair-use arguments of the vendors of these models, I don't have a lot of faith they would play hard-ball with a litigant who has code that has been established to have been generated, but who argues sufficiency from a "trust me, bro" perspective. (IANAL either, of course)

I would *love* to be wrong about that though, and I'm glad that the Copyright Office has drawn a clear line in the sand on the general matter (and wish more people in tech had read either the publications themselves, or this CRS summary of same)

@emma Oh yeah, shit's gonna get weird for a while and I think a lot of legislation going in during this administration as well as recent SCOTUS cases will need to be revisited. Ideally after also instituting laws around conflicts of interest with government officials that don't carve out exceptions for, oh I dunno, members of Congress, for example.

Basically, I want the different branches of the government to fight each other again rather than the different parties.