Artificial Intelligence and Copyright
Written by Florencia Gutiérrez and Mariel Chichisola
Society and technology are advancing exponentially. Artificial Intelligence (AI) is on its way to becoming a fundamental tool for addressing concerns and performing tasks, whether they are work-related, academic, or even for entertainment purposes.
What does artificial intelligence mean?
Artificial intelligence encompasses several different techniques. One of the most widely used techniques is machine learning, which relies on learning algorithms to identify economic trends or provide personalized recommendations. An algorithm is a precise set of sequential instructions or a methodical series of steps introduced by humans that can solve a problem, make decisions, or perform calculations[1]. So, one could say that machine learning AI systems are given a set of instructions or data in order to improve the results generated by the AI.
Now that we have a basic understanding of AI, let’s take a look at some of the most popular platforms currently available, their legal ramifications, and the products they generate. Some of the most popular AI platforms include ChatGPT and Dall-E, Stability.IA, Midjourney and Devianart.
What are these artificial intelligences, where do they come from, and what are their specific functions?
ChatGPT can answer any question from planning a trip to composing song lyrics in seconds and Dall-E, Stable Diffusion, Midjourney and Devianart generate images based on text descriptions.
One of the most widely used AIs is Stable Diffusion. This AI was developed by the company Stability IA and is open source, meaning that the source code of the software is publicly available for anyone to use, modify, and distribute it[2]. In addition, Stable Diffusion AI uses a machine learning system, which means that the more it is used, and the more information is fed into its system, the more it learns, and consequently, the better its results become.
From what images does Stable Diffusion create new images from?
According to a lawsuit filed by Getty Images in February 2023, the Stable Diffusion platform created by Stability IA has selected millions of images that were copied without prior authorization and used to train one or more versions of Stable Diffusion. See: “Getty Images (US), Inc. v. Stability AI, Inc. (1:23-cv-00135)”. Stability AI’s defense is based on dismissal for lack of subject matter jurisdiction. The defendant argues that the Delaware Court lacks jurisdiction over the Company’s UK entity and requests that the case be transferred to the Northern District of California, where a class action lawsuit has already been filed against Stability US and Stability UK, based on substantially identical facts and circumstances and raising identical legal issues.
Similarly, in January 2023, the trio of artists Sarah Andersen, Karla Ortiz and Kelly McKernan filed a class action lawsuit in the Federal Court of California (Northern District) – “Andersen v. Stability AI Ltd. (3:23-cv-00201)”. The lawsuit alleges that Stability IA illegally uses images from other creators to train its algorithm, even though the final output generated by the platform may be original. The defendant denies the claim, claiming that the artists cannot show a uniquely generated AI image that is substantially similar to the copyrighted works of the artists themselves. The defendant also argues that if the court were to rule on direct copyright infringement claims, it would have to analyze image by image to determine which works are protectable and which are considered fair use. Kristina Maude, an attorney with Dorsey & Whitney LLP, points out that “to determine fair use, a court should consider four factors: 1) the purpose and character of the use, 2) the nature of the copyrighted work, 3) the amount and substantiality of the portion used in relation to the whole, and 4) the effect of the use on the potential market for the work.”[3]
Given that these lawsuits are recent, having been initiated in this year 2023, there is currently no final judgment determining the copyright status of these types of artificial intelligences. In the case of “Getty Images (US), Inc. v. Stability AI, Inc. (1:23-cv-00135)”, it is estimated that the court will take 18 months for the court to render a decision.
However, in “Thaler v. Perlmutter (1:22-cv-01564)” Federal Judge Beryl Howell of the District of Columbia dismissed Stephen Thaler lawsuit challenging the government’s refusal to register works created by AI. Judge Howell found that copyright law has not been extended to protect works created by artificial intelligence operating without any human guidance. The ruling emphasizes that for copyright to exist, “human authorship is a fundamental requirement.”
In Addition, the rule on Copyright Registration in Works Containing Material Generated by Artificial Intelligence, published on copyright.gov on March 16, 2023, clarified that most works generated by AI are not eligible for copyright protection, but that AI-assisted materials may be eligible for protection in certain cases. Furthermore, according to Issue No. 1004, the United States Copyright Office launched a new initiative to examine copyright law and policy issues arising from the existence of artificial intelligence (AI), including the scope of copyright for works created using AI tools based on the use of copyrighted materials in their training. In this session, they discussed two specific topics: authorship in the context of generative artificial intelligence and the legality of “training” artificial intelligence platforms.
Conclusions:
Let’s imagine what happens in real life:an artist draws inspiration from another artist’s artwork. In that sense, that inspiration might not be considered a copyright infringement because it’s just inspiration. The idea or image that inspired the artist stays in the artist’s mind, shaping the creation of a new work. Now, applying this same concept from the realm of artificial intelligence, could it be argued that AI only uses these images as a basis for “inspiration”? Can AI be “inspired” or are its creations simply the result of combining existing creations of others?
Clearly, legislators face significant challenges when it comes to regulating artificial intelligence and its creations. Even the judiciary has a great responsibility in addressing issues related to AI-generated images. The rest of the world will be watching, as the cases mentioned above will provide some of the first signals of how the U.S. legal system may view AI and copyright infringement claims.
It is important that the algorithms behind these intelligences are designed and trained ethically and transparently, without undermining the contributions of great artists who shape the culture of nations and the world at large. But where is the line? We believe this is an fascinating discussion worth having in order to achieve the desired balance between human creation and ethics on the one hand, and innovation and technology on the other.
[1] Corvalán, Juan G. (2018). Inteligencia Artificial y derechos humanos. Microjuris online, cita: MJ-DOC-13757-AR||MJD13757.
[2] Camila Garcés S. Open Source, el próximo rey de la IA (2023). Recuperado de https://thevalley.es/open-source-el-proximo-rey-de-la-ia/
[3] Kristina Maude. Artists v. AI Images (2023). Recuperado de https://www.jdsupra.com/legalnews/artists-v-ai-images-2174061/
More information: Florencia Gutiérrez and Mariel Chichisola
