The Warhol Copyright Conundrum and the Future of AI Creativity

At the heart of the AI copyright debate lies questions of fair use. In Andy Warhol Foundation for the Visual Arts, Inc v Goldsmith, the U.S. Supreme Court made it more difficult to create content by imitating and amplifying existing artistic works. Justice Sotomayor, writing for the majority, held that The Warhol Foundation infringed the copyright of photographer Lynn Goldsmith when it licensed the portrait titled “Orange Prince” to Condé Nast in 2016. 

In 1984, Vanity Fair licensed a black-and-white photograph taken by Goldsmith so that Warhol could illustrate an article about the song “Purple Rain”. Warhol colourized, cropped, traced and exaggerated silkscreened versions of the photograph. Goldsmith claimed that The Warhol Foundation infringed her copyright by distributing Warhol’s variations of her photograph, which included the iconic orange version appearing on the cover of Vanity Fair in 2016. As recently reported by The Conversation, Warhol’s work was then sold to collectors and galleries and even sent to the Andy Warhol Museum in Pittsburgh. 

The U.S. Supreme Court held that The Warhol Foundation should have obtained permission from Goldsmith and included her in the deal. While the decision first reads like a win for independent creators “even against famous artists”, it raises concerns about the quality of artwork moving forward. Various legal experts have highlighted the potential for the Court’s ruling to “stifle a lot of creativity”, as noted by Artnet. Others have painted the decision as a “major victory” for copyright protection and commended its “reigning in of the copyright law’s fair use doctrine”. 

The recent judgment suggests that it could now be considered unlawful for artists to mimic existing works if the outcome is intended for a comparable platform, such as magazine covers and gallery exhibitions. According to the dissenting opinion, the majority forsakes the law’s historical emphasis on fostering creativity by imposing requirements for authorization and compensation. What does this mean for generative AI platforms that are trained on vast databases of human-made works? Only time will tell.