"I'm Afraid I Can't Copyright That, Dave" - Nutella and Machine-Authored Creative Works

If, like many others, you are a fan of Nutella, the chocolate-hazelnut spread that has become so popular that there is a "World Nutella Day" and at least one restaurant in the U.S. devoted entirely to Nutella-based dishes, you may have seen it in the news recently regarding its new advertising campaign in Italy. The campaign, "Nutella Unica" put approximately seven million jars of Nutella on the shelves in Italian stores, each with its own unique label design. Consumers snapped them up in just one month, and Nutella is considering expanding the idea to other countries.

But how did Nutella accomplish the creation and manufacture of so many unique labels? Unsurprisingly, (though perhaps disappointingly) they did not hire an army of artists and turn them loose, nor was this was an elaborate "design your own Nutella label" campaign reminiscent of "customized" Coca-Cola cans. Perhaps unsurprisingly to anyone paying attention to the Internet space at this point, the unique labels were the output of a computer algorithm, the parameters of which had been tuned to allow any pattern or combination of colors, but to never repeat itself. The resulting labels are being described by the ad agency responsible as "like a piece of art."

Hearing these algorithmic labels described as art got me thinking. If the labels are unique art, then it seems reasonable to suppose that they are copyrightable expression. This then presents the question of who owns the copyrights in the labels, which opens up a wide range of related questions. Is it Nutella? Is it the computer programmer(s) who designed the algorithm? Is it the purchaser of a given jar? Some kind of joint work? Should no copyright be assigned at all? Should the law require that copyright ownership for algorithmic outputs be addressed by work-for-hire frameworks? Whatever the answers, why? (And what if a monkey gets involved? Or a bolt of lightning strikes outside the factory while the labels are being printed?) It's likely not an issue for the Nutella labels, since I have a hard time imagining that anyone will want to make copies of any of them, but the underlying questions are powerful ones.

Although most people have a strong intuitive sense one way or another about copyright in machine outputs, the primary questions - whether copyright should exist, and if so, who should have it - are not easy ones to answer, because most of the primary theoretical justifications for copyright no longer apply when a computer program is the author, or do so only imperfectly.

In fact, the question of copyright for things created by machines has been debated since at least the advent of the modern Internet, if not well before, and continues to be a topic of legal scholarship. For example, a court held earlier this year that copyright protection afforded to a computer program did not extend to the program's output. However, the case concerned a fairly limited fact pattern, so although the court described questions of copyright in automated outputs as the most interesting aspect of the case, this particular ruling won't resolve any of the larger questions about copyright for machine authors/creators.

I've no doubt that as machine learning, algorithmic processes and AI generally move further and furthe