AI-Generated Works Can’t Be Copyrighted, Reaffirms US Copyright Office

The US Copyright Office has released a new report that clarifies and reaffirms their stance on AI-generated works, essentially stating that AI-generate works can’t be copyrighted.

This has become a huge issue, as there’s not just been an explosion of AI-generated text, images and audio, but it’s increasingly being used commercially in web content, Spotify playlists and other places where companies are trying to save costs by replacing people with computers.

The Office started their inquiry in August 2023, formally seeking public input on the full range of copyright issues that have been raised:

“In response, we received more than 10,000 comments representing a broad range of perspectives, including from authors and composers, performers and artists, publishers and producers, lawyers and academics, technology companies, libraries, sports leagues, trade groups and public interest organizations, and even a class of middle school students. Comments came from all 50 states and from 67 countries. That valuable and extensive input, supplemented by additional Office research and information received from other agencies, forms the basis for the discussion and recommendations in this Report.”

The Report Conclusion states that works may be copyrighted when AI is used as a tool, but not when the work was generated by an AI work from prompts:

“Based on the fundamental principles of copyright, the current state of fast-evolving technology, and the information received in response to the NOI, the Copyright Office concludes that existing legal doctrines are adequate and appropriate to resolve questions of copyrightability.

Copyright law has long adapted to new technology and can enable case-by-case determinations as to whether AI-generated outputs reflect sufficient human contribution to warrant copyright protection. As described above, in many circumstances these outputs will be copyrightable in whole or in part—where AI is used as a tool, and where a human has been able to determine the expressive elements they contain. Prompts alone, however, at this stage are unlikely to satisfy those requirements. The Office continues to monitor technological and legal developments to evaluate any need for a different approach.

The Office will provide ongoing assistance to the public on the copyrightability issues related to generative AI, including by issuing additional registration guidance and updating the relevant sections of the Compendium of U.S. Copyright Office Practices. In doing so, we will rely on the comments received in response to the NOI, judicial developments, and other relevant input.”

This is really interesting guidance for musicians, because it seems to make clear that:

  • AI-generated works cannot be copyrighted, making them public domain; and
  • Human-generated works that make use of AI tools can be copyrighted.

This has some interesting implications for musicians. It seems to suggest that all AI-generated music is public domain, making it potentially a huge new source of audio for musicians to sample, transform and use in original, human-generated music.

And it clarifies that music made with the use of AI-powered tools – such as the AI-powered beats of hardware tools like the Mutable Instruments Grids, or AI-powered stem separation – can be copyrighted, as long as the creative work is human-authored.

Do you have concerns about the rise of AI tools for music and AI-generated music? Check out the Report and let us know what you think!

Photo Credit: Public domain image of a cat, with human hands, reading a newspaper taken from the above US Copyright Report.

21 thoughts on “AI-Generated Works Can’t Be Copyrighted, Reaffirms US Copyright Office

    1. john – this is a sentiment that many of us probably share, but the genie’s out of the bottle on this.

      My personal take is that creative people need to explore the idea that these AI startups are spending billions of dollars on creating tools that are designed to replace us, so we should be taking advantage of the fact that this is public domain content.

      For example, if there’s ever a AI-generated song that gets any traction, musicians should be able to create copyrighted derivative works.

      As a point of reference, the longest-running US Number One of all-time is Lil Nas X’s ‘Old Town Road’, which is based on a sample from NIN’s Ghosts, released under a Creative Commons Attribution-Noncommercial-Share Alike. Because of the licensing, the sample had to be cleared by NIN, and NIN’s Reznor & Ross get their credit and a cut.

      With AI-generated public domain works, this should not be necessary.

      1. synthhead – as a mere technology, i don’t have a problem with artificial intelligence. it can prove useful in science, research, image recognition, astronomy, maybe even friggin particle physics, and gumball machines!

        what i have a problem with is:
        1. lumping every technique related to AI *as* AI. we’re not even close to a General Artifical Intelligence in these recent contraptions. even Algorithmic Intelligence is beyond us – (yeah we gots progress). they’re no more Intelligent than Expert Systems or the result of a long list of logic oriented programming languages. this isall marketing-up a basic algorithm dressed up as “A.I.”
        2. it’s being used to con people, and impact folks in a negative way.
        3. it’s ugly. all the images are unnerving, as if seen through a souless eyeball (like the alien eyeball on a stalk in the original World of the Worlds Ptuu Pttuuu ptttu!!) that can’t recognize humanity as more than shuffle-exchange of a corrupt pointer to a tree data structure on a stack.

        evolution took millions of years to get to the human brain, let’s not sell it short so quickly. lol.

        to me, the whole creative argument thing is subjective to you ‘music’ folks. as are many of the silly arguments here. :0) but, let’s not start another.

        1. john – there seems to be a lot of denial or confusion in your comments.

          1) AI and General AI are different things. AI is already probably better than you at math, chess, art and many other activities, and is certainly better than the average person. This is a big deal.

          2) There definitely is an ‘AI’ hype, but that doesn’t mean that AI is not hugely significant. Remember the dotcom hype? It was a bubble, but it also turned out to be significant as hell.

          3) People like AI-generated art better than human-created art already: (https://maxread.substack.com/p/people-prefer-ai-art-because-people)

          AI companies are coming for music already, too. Spotify is slipping AI-generated music into your playlists, because they can keep you listening without having to pay any musicians.

    2. Asking for AI to ‘just go away’ is like asking for electricity or running water to go away.

      Unless you’re Amish, you should be thinking about how you want to use AI.

  1. Its like a user of a copier occupying copyrights of copies it just copied because he pressed the copy button.

    Just my two cents and maybe for clarification: MI Grids has no AI involved while the module itself is running! The patterns in the module (Arduino microcontroller) are hard coded! Maybe Emilie was using AI to get the patterns, but there is no AI in the module inside.

    1. For Grids, machine learning was used to train on actual drum patterns and used to create the grid of options available on the module.

      Grids as an example of the extreme end of AI applications that should NOT be confused with ‘AI generated’ music. At the other end, there are now applications that just spit out bad music.

  2. Calling mutable instruments grid “AI” only displays a complete lack of comprehension of the topic.

    The grids sequencer is algorithmic and not learning-based: Grids doesn’t “learn” from input or adapt over time like AI can. Instead, it selects rhythms from a predefined dataset of drum patterns and modifies them using probability.

    So no need to add more hype to the AI bandwagon.

    1. Rob – The ‘grid’ in Grids was AI-generated using machine learning, so it sounds like you may just have a lack of knowledge of either the module or how machine learning works.

  3. Any time I see a buzzword-type trend appear, I automatically assume it’ll take a decade to prove it either useful or bogus. The cat picture isn’t even GOOD AI; the fingers and misplaced pipe are glaring. AI is proving useful in doing things like mass comparisons of possible new drugs, but when it “creates Art,” its like encouraging some sweet soul with Down’s to recreate the Mona Lisa. The smile just ain’t gonna be right.

    There are some applications of it that are colorful and interesting, but the copyright office is correct in defining it as not eligible for copyright. That should go to a human being, not an algorithm.

  4. nah man – you havent heard the end of this yet. Adobe has a new “AI replace” brush in the toolbar for Photoshop now, so you can brush away mistakes with AI… or brush a background into full view with AI, etc….

    so its much more complicated than you might think

  5. Yea, it seems like it might be difficult to prove whether or not AI produced a work., and therefore whether or not the work can or can’t be copyrighted.

    Regardless of the other dangers posed by new tech; AI & Crypto require silly amounts of electricity to run. Kind of not what humanity needs at the moment.

  6. AI-generated works cannot be copyrighted, making them public domain + 1 piano note = Human-generated works that make use of AI tools can be copyrighted. Profit. Correct?

  7. But how do you prove that a piece of music is generated by a text prompt? In other words, if somebody trains an AI model on your style of music, generates derivative music by it and then claiming copyrights against you on all streaming platforms, how do you protect yourself from these kind of scams?

  8. I’m not sure how they are going to draw a line and enforce this? It seems just like a temporary proclamation to appease worried people but it will never really work and the technology will evolve beyond the point of control?

    I’m reminded of when digital audio recording became possible at the consumer level with the DAT format and there was a big todo over trying to make it impossible to make perfect copies of commercial releases. Unfortunately any method proposed was either easily defeated or negatively affected the audio and the only thing the efforts did was to kill the DAT market and then eventually technology advanced and there was really no way to control copying any more.

  9. Lol. Phase-shift a wav a fraction and the algorithms can’t tell the difference who created it, least of all “The Office”

Leave a Reply

Your email address will not be published. Required fields are marked *