The Federal Circuit has affirmed the denial of copyright protection to an AI-generated image on the grounds that copyright requires a human author. As far as I know this was the expected outcome; I certainly think it’s correct. I talked about the case a bit and made a couple of policy arguments against AI copyright here, when the lower-court ruling came out.
The appellate decision lists several reasons AI cannot be an author: (1) copyright authorship is premised on the capacity to hold property, which AI cannot; (2) copyright duration is tied to the author’s lifespan; (3) copyright includes inheritance conditions, and machines don’t have heirs; (4) Copyright transfer requires a signature, but “machines lack signatures, as well as the legal capacity to provide an authenticating signature;” (5) authors are protected regardless of their “nationality or domicile,” but machines have neither; (6) authors have intentions whereas “Machines lack minds and do not intend anything;” (7) when the copyright act does talk about machines, it always talks about them as tools.
As the court summarizes:
“All of these statutory provisions collectively identify an “author” as a human being. Machines do not have property, traditional human lifespans, family members, domiciles, nationalities, mentes reae, or signatures. By contrast, reading the Copyright Act to require human authorship comports with the statute’s text, structure, and design because humans have all the attributes the Copyright Act treats authors as possessing. The human-authorship requirement, in short, eliminates the need to pound a square peg into a textual round hole by attributing unprecedented and mismatched meanings to common words in the Copyright Act.” (12)
The court notes that these are not necessary conditions, citing for example the inability of nineteenth-century women to own copyright, even though they could author works that were eligible for copyright protection. Nonetheless, “the current Copyright Act’s text, taken as a whole, is best read as making humanity a necessary condition for authorship under the Copyright Act” (13).
The rest of the opinion works through the regulatory development of the statutory text by the copyright office and refutes Thaler’s position. I thought this part was refreshing: “Dr. Thaler also argues that the human-authorship requirement wrongly prevents copyright law from protecting works made with artificial intelligence;” but “the Supreme Court has long held that copyright law is intended to benefit the public, not authors” (18). This is really important, and recent jurisprudence has made it harder to recognize the public-facing intention of IP law. I wrote about this tend toward privatization in my IP book, and argued that there was a clear public biopolitical intent behind earlier copyright, especially as it was discussed and around the U.S. Constitution drafting and by people like J.S. Mill and MacCaulay in the 19c.
The Court notes that copyright law has evolved with technology and that current doctrine doesn’t preclude protecting works made by AI – the AI just can’t be the “author.” It has to be a tool used by a human author, with resulting “line-drawing disagreements over how much artificial intelligence contributed to a particular human author’s work” (19).
Those of course aren’t relevant here, because Thaler said the AI was the “sole” author of the work. It seems to me that Thaler’s model is a work for hire: if I hire somebody to paint a picture and tell them what it should be of, they can create it and I can become the copyright owner of a work for hire. That of course requires that the work be the product of a painter. So Thaler has argued himself into the least tenable way of thinking about AI and copyright. Everybody who’s tried to create an image using those text-to-image tools knows that the prompting is a lot of the effort. So I would expect the line-drawing cases to continue for a long time, and we may well see efforts to get IP protection for certain prompt-engineering strategies or sequences.
I still think the a big part of the issue is there being too much copyright protection in the world. I just asked Gemini to “draw a picture of a chatbot thinking about whether it can get copyright for an image it makes,” and it produced the following:
I then said “now draw a different picture of the same thing” and it produced this:
I spent like 8 seconds doing that, and if copyright can inhere in this, I’ve created two copyrighted images in those 8 seconds.
The picture at the top – my third copyrighted work of the last couple of minutes if AI is copyrightable – is from a third prompt, “now draw a picture of a chatbot that's sad about not being able to get copyright for its images.”
Recent Comments