For copyright purposes, how can AI-generated content (eg.lyrics) be proven as such rather than human-generated?

Question

Most answers I have seen to the question of whether AI-generated content is copyrightable cite law in the US and other countries limiting copyright to human-created content. However all the answers I have seen fail to address the question of how a government copyright office and then potentially later a rights holder, copyright enforcement company, or court would be able to prove it was AI-generated.

AI platforms are often vague about retention and/or use of user input and AI output, and it is not clear to me how accessible this data would be to government copyright offices, courts, and copyright enforcement companies for purposes of determining whether a particular lyric were AI generated.

In some cases, already copyrighted content may appear verbatim in AI output, and to the extent this can be easy recognized as an existing work this would make the AI authorship question moot. However most of the time this will not be the case, especially where AI only enhances a user's original work rather than composing something from scratch or with minimal prompts.

Answer 1

In long-run, I would say it is a lost battle. Sooner or later AI can generate content that is indistinguishable from real human's work, just like a handsewn sweater versus a factory-produced one.

Even if all AI platforms fully comply in every way they can, that does not stop small work groups or even individuals to create their own AI tools. And it only gets easier and easier as technology advances.

I have seen some artists being asked for proofs that they did not use any AI tools, to which they complied by showcasing how they created the work (draft, diary, elaboration of thoughts and motivations). However, given sufficient demand, there will be AI which can generate 'proofs' (or simply fake some paperwork). To some point these proofs will become so real that fools everyone, again.

So I am on the pessimistic side. The most feasible solution I can think of is to fight magic with magic: train an AI detective. Or we can do the old-fashion way: have a couple witnesses to monitor every moment an artwork is created. But alas, human lies a lot too.

Answer 2

Although often today there is a stylistic "tell" to AI generated content, whuch might raise suspicion (for example it's still relatively easy to spot cut&paste from ChatGPT in Stack Overflow answers, and the mangled fingers and weird not-quite-sure-what-that-is parts of generated artworks), there is no reason to expect that to always be the case. It is likely to be a technical impossibility to detect high quality AI generations, as distinct from human creative works, purely from their content, in the not too distant future.

In humans, the creative process takes time and effort, and will leave evidence. For instance a writer will have notes, early drafts and also have provenance for other works they have made. They will be able to explain how they got the idea for a creative work, and at least some of the decisions they made in order to arrive at a final published version.

In contrast, an AI generated work without human input is generated complete without any other artifacts - the producer of it will often not comprehend it fully, and won't have any evidence of the work they did. They may also not have a history of producing similar works, and be unable to demonstrate any skill with the type of content they claim to have made, when challenged under controlled conditions.

This does leave more subtle interactions, where a skilled artist uses AI to augment their work, where such challenges would not be able to tell. But this is not very different to difficult human vs human "art theft" cases, where a lot of the decision is not technical, but based on statements made by plaintiffs and defendents in court. There will be some success rate due to the fact that most people will stick to a variation of the truth as opposed to blatantly lie about things they did. There may also be witnesses to behaviour that corroborate one side or another in a dispute. This will not be part of any technical test, but then there are plenty of other legal issues where there isn't a simple technical test, so that doesn't break anything.