Inventions cannot be Inventors, Yet!!

Original article was published by Harshit Singh Jadoun on Artificial Intelligence on Medium

Inventions cannot be Inventors, Yet!!

The article is written by Ojasvi Mishra and Aditya Vikram Singh, students of the Institute of Law Nirma University, Ahmedabad.

Photo by Hitesh Choudhary on Unsplash

With growing instances of Artificial Intelligence (“A.I.”) being employed on almost every aspect of the creation across various industries, we see increasing number of important developments in technology and business being driven out/created by these machines. Undoubtedly large volumes of training data and funds are being invested in boosting A.I.’s growth. As an inevitable result, A.I. systems are now taking on tasks formerly performed by and associated only with humans, including creation of intellectual property. Some recent examples of IPs created by A.I. machines are The Next Rembrandt” project where the A.I. produced paintings replicating the artist’s subject matter and style. Some other examples are, the Emily Howell or Bot Dylan computer programs which have composed music, or an A.I.-written novel The Day A Computer Writes A Novel”.

There have been some deliberations and some legislative premise developed over the years, related to A.I.-generated works and copyright law. In 1988, UK for the first time came up with a policy to provide explicit copyright protection for A.I. or “computer-generated” works. As per this, when an otherwise copyrightable work is created but no natural person in those circumstances qualifies as an author, the “producer” of the work is deemed to be the author. On the other hand, the US Copyright Office has taken the opposite approach, applying its “human authorship policy” since 1973.

Patentability of A.I.-generated inventions

However, there have been no solid legal framework on patentability, of A.I.-generated inventions. Almost all the jurisdictions require patent applications to mention an inventor who is a natural person. Recently, in August last year, there were 2 international patent filings for inventions generated autonomously by an A.I. system called DABUS. These applications listed the A.I. as the inventor and the A.I.’s owner as the patent applicant and the prospective owner of any issued patents. Both the European Patent Office (EPO) and United Kingdom Intellectual Property Office (UKIPO) evaluated these applications on their merits and held that these applications met the requirements of patentability to the extent possible prior to the publication of the applications but both these institutions concluded that DABUS cannot be accepted as an inventor, because the designation of a natural inventor is compulsory, as there are legal consequences associated with registration of new inventions and related patents. A.I. systems or machines have not yet been given the recognition of a legal personality and thereby do not hold any independent rights. Moreover, just giving a name to a machine would not be enough to satisfy the first requirement mentioned above.

IPO also pointed out another significant issue that, all artificial intelligence machines have an owner or creator, but in the contemporary legal structure there is deficiency of legal procedures regarding transfer of rights from machines to their human owner.

Earlier this year, The US Patent and Trademark Office (USPTO) also held that A.I. machine systems cannot be titled as an inventor in a patent. In laying down this rule USPTO reasoned that, the language used by the legislators in drafting US patent law constantly refers to inventors using humanlike words such as “whoever” and pronouns like “himself” and “herself.” The USPTO also cited some rulings of the Federal Circuit wherein it held that only natural persons can be referred to be inventors, for example, in University of Utah v. Max-Planck-Gesellschaft zur Forderung der Wissenschaften, the Federal Circuit rejected the claim that corporations or sovereigns could be inventors.

In its decision, the USPTO also deliberated upon various public policy arguments offered by the applicants, such as; allowing an A.I. system to be registered as an inventor would not just incentivize innovation but also reduce the improper naming of inventors, and thus it would promote public notice of an invention’s actual creator. However, USTPO considered these arguments to be insufficient to surpass the literal interpretation of U.S. patent laws as read by the courts over the years.

While this decision clearly ruled out that under the existing law an A.I. system can’t be an inventor, it leaves mist of ambiguity on many relevant questions. For instance, this question remains unanswered that what legal protections are available to owners of inventions conceived with the assistance of A.I., and if available, up to what extent? Similarly, as the decision was specifically in the milieu of patents, it does not bring clarity on the position of other forms of IP created by A.I. without human assistance.

Consequently, this decision in some way fueled the long running debate on the implications of A.I. on traditional perceptions of IP rights under U.S. law while putting forward the big question that, whether the current laws are sufficient or there is a need of revamping this legal premise to address new and emerging technologies.


Though it’s true that an A.I. would not get motivated by holding a patent, undoubtedly, it will inspire and encourage those people who own, develop and use A.I. thereby, resulting in more innovation for society in the long run. Other than providing protection for AI-made inventions, there are also moral reasons for listing A.I. as the inventor when it is actually inventing. If we allow a person to be listed as an inventor for an invention functionally created by A.I. may although not be unfair to an AI, since it has no interest in such acknowledgement, but allowing people to take credit of inventions they have not made, would disparage human inventor ship. It would put the work of a person who just commands an AI system to solve a problem on the same pedestal with those who work hard to actually create something new.

As of now, AI systems dearth legal rights and thus are not eligible to own property. Thus, it becomes even more important that the owner of the AI should own any patents on inventions made by its AI, according to the general principles of property law as well as fundamental principles that apply to other areas of intellectual property like trade-secret protections. However, as on date there exist no obvious benefits to changing laws to allow AI ownership and there would be significant costs in doing so, it may take a few years to overturn this legal scenario.