Picasso? Monet? Renoir? Bah, who needs them anymore?
By Joseph R. Patrell
Congress passed 17 U.S.C. § 101 in 1976, which is commonly known as the Copyright Act (“The Act”).[1] Among other things, the Act grants exclusive protection to copyright owners to exclude others from reproducing and displaying the copyrighted item.[2]
Under the Act, one of the requirements for copyright protection is Initial Ownership.[3] As defined vaguely under § 201(a), copyright protection “vests with the initially in the author, or authors of the work.”[4] Additionally, that protection allows authors to recover monetary damages and receive injunctions against infringers.[5] One of the theoretical underpinnings of Copyright law is Lockean Labor Theory.[6] Under this theory, creators are granted the rights to their intellectual property because of the physical and intellectual labor that went into creating the expression.[7]
Congress could not have predicted the rise of the internet in the 1970s when creating the Act. During the 1990s, when the internet became commercially available, many new and unexpected problems for copyright owners were created, especially in protecting their work from infringers.[8] As a past JOLT article from 1995 mentions, with the lack of clarity in the Act on protecting intellectual property during the rise of the internet, a “full-scale battle of interest groups will take place concerning any attempt to amend the Copyright Act.”[9]
Now, again, the Act comes to more crossroads where Congress has been ambiguous about what constitutes an “author.” When an artificial intelligence (“AI”) machine makes the creations, is that considered an author? How do we protect artists’ creations from AI machines scraping the internet for creation learning? And, when artists do allow AI machines to use their creations for learning, should the artist receive royalties every time or one time?
Two recent events involving these questions are playing out currently, with no concrete guidance from Congress. The first event involves Shutterstock, the online stock image company. On October 25, 2022, Shutterstock announced that they would partner with OpenAI’s DALL-E2 AI generator to sell AI creations on their platform.[10] At a high level, DALL-E2 generates and manipulates images from text prompts by a user.[11] For DALL-E2 to work, it must be trained with millions of images, thus allowing the machine to generate AI creations from any random assortment of words that users decide to put into the machine.[12] In this announcement, Shutterstock laid out its plan for paying artists whose work is used by DALL-E2.[13] Shutterstock’s plan is to launch a new fund to compensate those creators whose work has been used.[14] But what’s interesting from this announcement is Shutterstock’s choice to create a new payment model, breaking away from the traditional licensing payments creators have received in the past.[15]
This raises the question of whether or not creators are getting the short end of the stick. Is Shutterstock theoretically able to avoid the traditional royalties owed to creators because the AI machine only needs to access the creation one time? From that single use of the creation, the machine can produce unlimited amounts of generated images. Furthermore, critics have also raised concerns about AI machines scrubbing the internet for images to learn from without authorization from the creators.[16] When an AI generator uses an unauthorized image for training, would this action fall under copyright infringement or fair use?
It is also challenging to determine who the author is when AI-generated work is created. In 2018, Stephen Thaler filed an application to register an AI-generated creation.[17] In the application, he listed the AI computer as the author of the work.[18] The Copyright Office denied his application due to the lack of human authorship.[19] After internal agency actions, Mr. Thaler filed suit claiming the Act does not explicitly state the requirement of “human authorship” but only requires “original works of authorship.”[20] Further, he argues that because corporations can be considered authors, the same logic should apply to AI-generated creations.[21] Traditionally, the Copyright Office has been strict on the human authorship requirement. The lack of clarity in the current text of the Act and the rise of AI creations will likely only continue to pose issues for both Copyright owners and those seeking registration on AI-created works.
In sum, there is a magnitude of questions with no clear answers currently. And while this problem is novel for the legal system, we should emphasize compensating the artists whose creations allow the AI machines to learn in the first place.
Image Source: Joseph R. Patrell
[1] 17 U.S.C. §§ 101-1511.
[2] 17 U.S.C. § 106.
[3] 17 U.S.C. § 201(a)
[4] Id.
[5] 17 U.S.C. §§ 501-513
[6] Lydia Loren & Joseph Miller, Intellectual Property Law: Cases & Materials 325-327 (7th ed. 2021).
[7] Id.
[8] I. Trotter Hardy, Contracts, Copyright and Preemption in a Digital World, 1 RICH. J.L. & TECH. 2 (1995) http://www.richmond.edu/jolt/v1i1/hardy.html.
[9] Id.
[10] Press Release, Shutterstock, SHUTTERSTOCK PARTNERS WITH OPENAI AND LEADS THE WAY TO BRING AI-GENERATED CONTENT TO ALL (Oct. 25, 2022), https://www.shutterstock.com/press/20435.
[11] OpenAI, https://openai.com/dall-e-2/ (last visited Dec. 2, 2022).
[12] Id.
[13] Shutterstock, supra note 10
[14] Id.
[15] Earnings and Payments, Shutterstock, https://support.submit.shutterstock.com/s/article/How-much-will-I-be-paid-as-a-contributor-to-Shutterstock?language=en_US
[16] James Vincent, Shutterstock will start selling AI-generated stock imagery with help from OpenAI, The Verge (Oct. 25, 2022), https://www.theverge.com/2022/10/25/23422359/shutterstock-ai-generated-art-openai-dall-e-partnership-contributors-fund-reimbursement.
[17] Complaint at 1, Thaler v. Perlmutter, 1:22-cv-01564 (D.C. Cir. Jun. 2, 2022).
[18] Id.
[19] Id.
[20] Id.