Patents and AI
Patents and AI

Explore AI Corner

Short Guides about Patents and AI

AI inventorship

PA1 Quick Guide- AI Inventorship E1.svg

  • Nowadays, AI can analyze large amounts of data, find patterns, and combine the extracted information to generate something new. This process has been used to create artworks and inventions, making an exciting development in human history. However, it also raises important questions in the fields of copyright and patent law. Let’s see why.
  • The outputs generated with the use of an AI model can be used in an invention or could even be considered inventions in themselves. This has led to people citing AI as an inventor in patent applications, as in the DABUS case before the US Patent and Trademark Office (USPTO). However, most patent offices have rejected these applications because they generally require the inventor to be a natural person. 
  • One reason behind the decision is that patents are meant to give inventors exclusive rights to their inventions, provide recognition, reward their ingenuity, and offer the possibility of a return on their investment. Through this, the grant of intellectual property seeks to incentivize innovation. 
  • However, AI does not need those incentives to invent; it creates inventions because it is programmed to do so, not because of the benefits that it might derive from the invention.
  • Those AI-generated inventions are still desirable for the innovation ecosystem but cannot be attributed to or owned by an AI. 

PA1 Quick Guide- AI Inventorship E3.svg

  • There is a debate on whether a person should be entitled to inventorship to introduce an invention generated by AI, particularly when they have invested resources in developing the AI system. Some scholars argue that if an inventor creates an AI system that generates an invention, that inventor should be eligible for inventorship of that invention. 
  • This discussion may seem premature, as AI is currently just a tool in the invention process, like other tools that facilitate human discovery and technological progress. But it is necessary to think ahead and pave the way for when regulation is needed.
  • Some jurisdictions have started pointing out that the level of AI's contribution to the inventive process would be a key factor in determining patentability. This includes a spectrum ranging from AI-assisted inventions, which still involve a fair amount of human input, to AI-generated inventions, where AI systems can independently generate solutions without detailed human direction. The challenge lies in determining where on this spectrum an invention falls and whether the human input is significant enough to warrant inventorship. 
  • These points highlight that the concept of AI inventorship is complex and multifaceted, with legal, ethical, and practical implications that require further consideration and possibly even a reinterpretation of current patentability requirements.

AI1 Quick Guide - How AI Works H1.svg

References

AI and the inventive step

AI2 Quick Guide – Machine Learning and Neural Networks G2.svg

  • The emergence of AI is causing tremors in many industries, including patent law. AI is testing the traditional patentability criteria—novelty, inventive step, and industrial applicability.
  • The "inventive step" or "non-obviousness" requirement prevents the patenting of trivial advancements. It ensures that an invention is not an obvious modification or combination of existing technology from the perspective of a person skilled in the art, promoting true innovation. 
  • Inventive step is assessed from the perspective of a “person having ordinary skill in the art” (PHOSITA). This hypothetical person with ordinary creativity, not an automaton, has access to the same tools, skills, and knowledge base as others in the field. The level of skill of the PHOSITA is critical in determining obviousness. 
  • In Europe, the problem-solution approach is used to assess inventive step. This approach involves identifying the closest prior art, determining the objective technical problem to be solved, and assessing whether the invention would have been obvious to a PHOSITA. The objective technical problem should be one a skilled person would wish to solve, knowing only the prior art. The question is not whether a PHOSITA could have made the invention but whether they would have done so in expectation of an improvement.
  • When AI is used for inventive purposes, it can be difficult or almost impossible to understand how AI found the solution to the problem. This is problematic because it is necessary to disclose enough information in the patent so that a PHOSITA can understand and reproduce the technical teaching of the invention. 

PA2 Quick Guide- AI and Inventive Step D3.svg

  • A key issue is whether the PHOSITA should be considered as someone using AI models. If a PHOSITA is assumed to have access to AI, this will raise the standard for non-obviousness because she/he would have access to broader information and to combinations of inventions that humans have not necessarily tested.
  • The Board of Appeals of the European Patent Office (EPO) has noted that the mere application of a known machine-learning technique to a particular field is a general trend in technology and may not be inventive (see case T 0161/18).
  • These discussions have led some scholars to suggest re-thinking the non-obviousness requirement in light of AI development. The discussion focuses on the adequate threshold, too high a threshold could prevent deserving inventions from being patented and lead to slower technological development, while too low a threshold could lead to a flood of patents that would hinder those skilled in the art. 

 

PA2 Quick Guide- AI and Inventive Step D1.svg

References

Research Papers on Patents and AI

Webinars on Patents and AI

Inventorship and AI: Defining the Human Role in Patent Law

This webinar, with Professor Noam Shemtov and Heli Pihlajamaa, explores the evolving landscape of inventorship under patent law, with a particular focus on the level and quality of human contribution required to qualify as an inventor in the context of AI-assisted inventions.

Podcast episodes

Listen on Spotify Listen on Spotify Listen on Apple Podcasts Listen on Apple Podcasts Listen on YouTube Listen on YouTube Listen on Amazon Listen on Amazon