Recent decisions around the world have begun to shed light on the question of whether works and inventions produced using GenAI systems are eligible for copyright and patent protection.
In this two part blog, we compare and contrast positions on this question in China, the US, the EU and the UK.
For Part 2, covering patent protection of AI-generated outputs, see here.
Copyright
China
In November 2023, the Beijing Internet Court (a first-instance court) ruled on the ownership of copyright in an image generated using an open-source AI image generator, Stable Diffusion. The plaintiff generated the image by inputting various positive and negative prompts into specific model packages, setting the desired parameters, and then adjusting the prompts and parameters in four rounds.
The plaintiff posted the final image on a social media platform. He then later discovered the image used in a post on a different platform, and sued for copyright infringement.
The Court held that there was copyright in the AI-generated image. The Court was of the view that Stable Diffusion was merely a tool (making an analogy to a camera). The Court held that the specific prompts and parameters chosen by the plaintiff provided the intellectual input required to create an original, copyrightable drawing, and that the copyright was owned by the plaintiff.
The awarded a fee of RMB 500 to the plaintiff (being the minimum sum of damages payable under the PRC Copyright Law).
US
In contrast, the United States Copyright Office (USCO) has both denied and cancelled copyright applications and registrations for works created using generative AI due - to lack of human authorship.
In February 2023, the USCO cancelled a copyright registration for the comic book Zarya of the Dawn, by Kristina Kashtanova, on the basis that the images in the work had been generated using Midjourney. Although Ms. Kashtanova argued that Midjourney served “merely as an assistive tool”, the USCO held that, because Midjourney’s specific output cannot be predicted by users and Midjourney exhibits a degree of autonomy, the use of Midjourney differs from other assistive tools.
While Ms. Kashtanova argued that the material was protectable because she expended significant time and effort, the USCO disagreed, reiterating that time, effort and expense has “no bearing on whether a work possesses the minimum creative spark required”. While Ms. Kashtanova’s selection and arrangement of the images were protectable as a compilation, the USCO found the images themselves were not protectable as copyright due to lack of human authorship.
However, the USCO did leave open the possibility that it may reach different conclusions for “other AI offerings that can generate expressive material [that] operate differently than Midjourney does”.
Later, in August 2023, the USCO denied a copyright application for Stephen Thaler’s (Dr Thaler) visual work A Recent Entrance to Paradise. The application listed an AI computer algorithm used by Dr. Thaler, the “Creativity Machine” as the author. The US District Court for the District of Columbia, in affirming the USCO’s denial of copyright, acknowledged the U.S. Copyright Act was intended to adapt to changing technology but held “human authorship is an essential part of a valid copyright claim” and therefore a work generated entirely by an artificial system is not eligible for copyright.
The USCO is currently inviting comment on various copyright issues relating to the use of AI, in particular copyrightability and registration.
UK
The issue of copyright protection for AI-generated output is yet to be adjudicated on in the UK. Under UK law, for a literary, dramatic, musical or artistic work to qualify for copyright protection, it must be ‘original’. The interpretation of “originality” under UK law has been influenced by EU legislation (including the InfoSoc and the Database Directives) and is considered to generally require that works are an author’s own intellectual creation (which would imply a requirement for “personhood”).
Assuming that UK law is found flexible enough to support a finding of copyright in an AI-generated work, the UK Copyright, Designs and Patents Act already provides that the author of a computer-generated work is considered to be the “person by whom the arrangements necessary for the creation of the work are undertaken”.
For works created by simple software programs, traditionally this has meant the human author who used the software to produce the work. However, it’s unclear how this would play out for an AI-generated work:
- Depending on the nature of the AI tool and its degree of autonomy, the ‘person’ making an arrangement to create the output could be an algorithm.
- Alternatively, if the creation of the work involves a very sophisticated prompting process, those ‘arrangements’ could be considered as being made by the prompter.
- Further alternatives are possible – for example, the individuals training and weighting a model could be seen as making the necessary ‘arrangements’.
EU
Under EU copyright law, a literary, dramatic, musical or artistic work must be ‘original’ to qualify for copyright protection. This requires that works are an author’s own intellectual creation and generally implies that only creations made by natural persons would qualify for copyright protection.
The question of whether copyright protection could be possible where an AI system has been used by a human author will likely depend on the specific circumstances of the case; in particular on whether the way the AI system has been used can be said to have to generated the user’s own personal intellectual creation (similar to a photographer taking a specific picture using a camera). On the other hand, if the work was conceived more or less randomly without input from a human author in adapting the parameters and prompts (which seems to be the view of the USCO in the cases above), in our view, an EU authority is unlikely to attribute copyright authorship to any human being.
