The US Copyright Office recently ruled again that artworks created by artificial intelligence are not protected by copyright, upholding its practice of not registering AI-generated art as a copyright work.
Copyright protection in the United States
Unlike in this country, it is possible to officially register copyright protection in the United States. It used to be mandatory to apply to the US Copyright Office in order to obtain copyright protection. When the USA joined the Berne Convention, registration was no longer required. Nevertheless, it is common practice in the USA to secure advantages such as proof of authorship or to obtain protection against the import of unlawful copies by customs. In addition, registration is a necessary prerequisite for filing a lawsuit in court.
AI-APP as co-author of a work of art?
The matter began in December 2021, when Ankit Sahni filed an application for registration for the work SURYAST[FR1], claiming two authors: himself as the author of the photograph “2-D artwork” and the “RAGHAV AI Painting App”. Ankit Sahni, a lawyer admitted in the USA, developed the AI-supported painting app RAGHAV himself.
His application was rejected on the grounds that the work “does not show the human authorship required for a copyright claim”. In the specific case, the human contribution to the overall work result could not be distinguished from the work contribution of the AI and separated from it. However, only a human creation is required for protectability.
Ankit Sahni argues that authorship cannot presuppose that a work is created entirely by a human author. Moreover, the AI he uses is merely a supporting software tool. It selects the original photo and determines the degree of stylisation. The program follows his instructions, namely to change the colours, shapes and style in a manner determined by him. The Office counters that the product created from the AI is at best a work derived from Sahni’s photograph, but that it does not itself contain sufficient human creation to justify registration as “copyrighted works”.
When analysing AI-generated material, the agency must determine when a human user can be considered the “creator” of the material. According to the Office’s March 2023 guidelines on AI-generated works, when examining an application for registration, it must answer the question of whether the computer technology used is merely an auxiliary instrument of an intellectual creation process already completed by a human being, or whether the traditional elements of the work creation (literary, artistic or musical expression) were conceived and executed by a machine (in which case copyright protection of the overall result is inconceivable). For such an analysis, it must be possible for the Office to clearly recognise from the registration applications what the AI-generated content is.
Success for “RAGHAV” in India and Canada
Ankit Sahni has had better luck in other countries: in September 2021, the Indian copyright authority recognised the RAGHAV Painting App as the author of the artwork SURYAST. In December 2021, Ankit Sahni also succeeded in registering RAGHAV as his co-author for SURYAST with the Canadian Intellectual Property Office (CIPO). This was the first time that machine-generated content was granted copyright in Canada.
This case clarifies two things:
- Who can claim a copyright and what rights derive from it is governed by the principle of territoriality, i.e. these questions can be assessed differently depending on the legal system. For those affected (creators and users of AI-generated products), the completely different copyright categorisation of AI-generated content around the world is unsatisfactory. This state of limbo will continue for a while in the absence of sufficient regulation. It therefore remains to be seen how the allocation of rights to artificial intelligence products will develop, and not only in the field of intellectual property.
- Sahni’s argument that AI is merely a technical means of realising the already completed human creative process is often heard, especially in the creative industry. However, it does not hold water as long as significant (if not all) creative steps are carried out by the machine, which is likely to be the case whenever results cannot be specified in every detail by humans. Artificial intelligence is not a clumsy executing technology, but is developed to produce results independently. This contradicts the principle of copyright law that only human creations are eligible for protection (remember the “monkey selfie” case).
Courts will soon have to decide,
- whether copyrights to AI-generated content arise and who owns them,
- whether rights are infringed by training AI with copyright-protected content or by using the content generated on this basis,
- who is liable for such (possible) infringements (operator of the AI? user of the AI? user of the products?)
Stay tuned. We will keep you up to date.