A brief history of patents
Frumkin (1947) writes that the first recorded patent pre-dates the industrial revolution, given in 1421 in Florence, Italy, to a famous architect named Brunelleschi, who invented a barge to transport marbles. In Canada, the Patent Act first came into effect in 1869. Patents were developed to incentivize innovation by giving exclusive rights to owners, but they also help disseminate technical knowledge that can be used in further inventions, and prevent wasteful innovation efforts by others working to create the same invention. Farre-Mensa et al. (2016) found that patent laws help startups and small businesses in innovation and growth, and that delays in patent examination can have major negative consequences for them and for their inventors (such as reducing firm growth, job creation, and innovation).
AI and the current patent framework
Afshar (2022) explains that patent laws require an oath and declaration that the applicant is the true inventor. Failure to include inventors, he notes, can render a patent unenforceable. There is legal precedent for this, with many patents being successfully challenged in court due to claims surrounding the inventors (whether the list of inventors was complete or accurate).
In the U.S., an invention is seen as a 2-step process:
- Conception of the idea or subject matter of the patent claims, which may be comprised of several claims; and
- Reduction of the idea to practice, or making a working example of the claimed invention.
U.S. law considers inventors as those involved only in the first step, with an exclusive focus on the “conception” of ideas that lead to inventions, or as Judge Graham (the presiding judge over the landmark patent case, Townsend V Smith 36 F.2d 292) put it, “the formation in the mind of the inventor of a definite and permanent idea of the complete and operative invention.”
An AI developed for a specific purpose, and given predictability that it would be able to generate results, could only be classified under the 2nd point, and hence not considered an inventor.
Re-drawing the line…
However, Tull and Miller (2018) explain that AI systems are becoming advanced enough to generate new inventions and develop their own code, with more complex machines expected in the coming few years. In such cases where an AI changes past its initial programming, and develops an unpredictable result, method, or technique, the question of who should be listed as the inventor becomes more complicated and unclear.
An AI machine that is capable of editing and updating its code, methods, optimization technique, and finds and solves its own problem, could present its owners, developers, and users with challenges to explain how they are the true inventors of its output. In other words, if humans were to be listed as inventors on a patent, while the AI was involved in the inventing process (without a human inventor’s input), this could violate the human inventor’s oath that they were the true (and only) inventors.
One way to solve this, is to list those who created the original systems as inventors. But once such considerations are given, the line gets even blurrier, as we’ve seen in Thaler’s case with the DABUS machine. For example, if a new medical or pharmaceutical discovery were to be made using an AI system that independently analyzed tens of thousands of drug formulations to find new unique compounds, should the AI programmers be credited as inventors? Or should medical experts who provide the AI machine with data be deemed the inventors?
Are there any benefits to allowing AI to be listed as inventors in patent applications?
Thaler’s lawyer, Ryan Abbott, argues that allowing AI inventorship would incentivize innovation, as solving issues surrounding patent inventorship for AI-developed inventions could boost further development of AI, and its use for general discoveries and inventions. He argues that humans should be listed as inventors only in 3 cases:
- If a person formulates a problem in a manner that requires inventive skills and instructs the AI to solve that problem.
- If a programmer specifically designs an AI to solve a specific problem, and skillfully selects training for the AI to be able to solve it.
- A person who recognizes the output of the AI when it suggests many options and the person uses inventive skill to choose an optimal solution.
However, Abbott claims, a person working on an obvious problem, a programmer who contributes to the AI’s general problem-solving capability but has no knowledge of the specific problem the AI is solving or its output, or a person who uses a straightforward and obvious AI output, should all not be listed as inventors. He argues that “listing an AI as an inventor is not a matter of providing rights to machines, but it would protect the moral rights of traditional human inventors and the integrity of the patent system.”
While laws that require a human to be listed as an inventor have, for many years, protected individuals working in corporations and ensured they would at least receive due credit for their work, these laws were not designed to take into account the possibility of novel inventions that could be developed by machines. Australian law recognized this already, with its legal framework allowing for AI inventors. Given the pace of development in AI systems in the 21st century, it could be time to clarify what it means to be “an inventor.”
Afshar, Mimi S. “Artificial Intelligence and Inventorship-Does the Patent Inventor Have to Be Human?” Hastings Sci. & Tech. LJ 13 (2022): 55–55.
Farre-Mensa, Joan, Deepak Hegde, and Alexander Ljungqvist. “The Bright Side of Patents,” 2016.
Fraser, Erica. “Computers as Inventors-Legal and Policy Implications of Artificial Intelligence on Patent Law.” SCRIPTed 13 (2016): 305–305.
Frumkin, Maximilian. “Early History of Patents for Invention.” Transactions of the Newcomen Society 26, no. 1 (1947): 47-56-47–56.
Gattari, Patrick G. “Determining Inventorship for US Patent Applications.” Intellectual Property & Technology Law Journal 17, no. 5 (2005): 16-19-16–19.
Li, Nick, and Tzeyi Koay. “Artificial Intelligence and Inventorship: An Australian Perspective.” Journal of Intellectual Property Law & Practice 15, no. 5 (2020): 399-404-399–404.
Maurer, Erik S. “An Economic Justification for a Broad Interpretation of Patentable Subject Matter.” Nw. UL Rev. 95 (2000): 1057–1057.
Tull, Susan Y., and Paula E. Miller. “Patenting Artificial Intelligence: Issues of Obviousness, Inventorship, and Patent Eligibility.” RAIL 1 (2018): 313–313.