raiyon—formerly known as DALL-E Mini—has captured the attention of the art and legal world in recent weeks, including this very blog in our first post on this fascinating service. The website, which touts itself as having the ability to produce “AI model drawing images from any prompt,” is undisputedly innovative and fun to use, yet raises myriad legal questions. To name just a few: Who authored the images, if the human user only input a text prompt and the AI did the rest? What if the resulting images are substantially similar to existing works? What if the resulting images use works belonging to others? Are the resulting images transformative enough to be fair use of preexisting works? Are the amounts of preexisting works used small enough to be considered de minimis? And most relevant to this post, who owns the images that result from the user’s text prompt?
Putting aside what could be prime material for Gen-Z law student exam questions on infringement and fair use, Craiyon just bumped the critical question of ownership to the forefront of all these inquiries, including who can make money off the images created by the site’s AI. According to WIRED, Craiyon recently announced that “users get full usage rights to commercialize the images they create with DALL-E, including the right to reprint, sell, and merchandise.” Assuming users can even have such commercialization rights if the works, themselves, may not be copyrightable (was there human authorship?), the company’s pronouncement begs more questions than it answers concerning users’ rights and potential legal exposure for the platform.
But then come the carve-outs. Individuals or entities “with less than 1 million United States dollars in annual revenue” can use the images generated on Craiyon “for the purpose of pursuing financial gain, provided that (1) a royalty of 20% is due on any revenues attributable to any blockchain transaction (such as selling NFT’s), and (2) we may revoke your right to use the Site or the Images at any time. Free commercial use requires you to attribute images to Craiyon…” Individuals and entities with more than a million dollars in annual revenue are instructed to negotiate with Craiyon separately for commercial use rights in the generated images.
These provisions sound a lot like license terms from a copyright owner to a content user. But how does the company have the right to dictate such conditions? Copyrighted works are de facto owned by their authors; ownership otherwise must be transferred by a specific written assignment or granted as a “work for hire” either by an employee for her employer or as a specially commissioned project for a specific subset of works, also subject to written agreements. A website or app doesn’t automatically own the content created by its users just because it supplies the necessary technology to create that content. For instance, programs like Adobe Photoshop do not own the content you make using their various artistic tools (although they will often require a non-exclusive license for certain uses necessary to operate their service and software).
With safe harbor status uncertain, Craiyon is taking a risk by allowing its users to commercialize the works generated by its AI and is essentially only relying on indemnification from individuals in the event of third-party claims, and query whether Craiyon could really even effectively enforce such provisions against individual users. Artists—photographers in particular—take their intellectual property rights very seriously, and it is more likely that a putative plaintiff will go after the platform as opposed to an average Joe trying to make a buck through a t-shirt side hustle.
Given how quickly technology develops and how it inevitably outpaces the law, Craiyon certainly will not be the first or only AI platform of its kind (take, for example, platforms like Jasper and Copymatic, who both automate the generation of written copy through AI). As more and more of these content-generating services emerge across creative disciplines, litigants and judges will have to determine how this new and fascinating technology fits into the decades-old existing legal frameworks or legislators will need to build new frameworks to accommodate the inevitably exponential grown of AI-based platforms.