AI and Patent Law: Navigating the New Frontier

On September 23, 2024, I had the opportunity to give a talk titled “Patent Law’s AI Dilemma: Innovation at the Crossroads” at the Louisiana State Bar Association’s “IP Trinity” CLE. The intersection of AI and patent law is a hot topic, and I’d like to share some key insights from my presentation with you.

Recent USPTO Guidance on AI

The USPTO has fairly regularly issued guidance on examination issues, mostly related to Section 101 “subject matter eligibility” issues raised by inventions in the fields of software and/or biology. They most recently issued new guidance on July 17, 2024, focusing on artificial intelligence in AI. Also, over the years, the USPTO has issued what I consider to be “teaching examples” of inventions plus discussions of how the subject matter eigibility rules apply to those examples. Notably, the USPTO issued its first new examples in about 5 years, with new examples 47, 48, and 49 being in the field of AI. To briefly summarize my thoughts on this new guidance:

1. AI-Assisted Inventions: The Guidance cites to prior USPTO guidance and explains that an invention cannot be patented if it is created solely by an AI system. Instead, to be eligible, a human must have made a “significant contribution” to the conception of the invention. Thus, only natural persons may be named as an inventor. this guidance appears to be consistent with recent Federal Circuit opinions. See, e.g., Thaler v. Vidal, 43 F.4th 1207 (Fed. Cir. 2022).
2. Section 101 and AI: The USPTO has provided examples of how they’re treating AI-related inventions under Section 101. They’ve emphasized that AI inventions are subject to the same subject matter eligibility criteria as other computer-implemented inventions. And, I think the best “teaching example” for AI is Example 47, which provides three claims. Example 47, Claim 1 is fairly straightforwardly eligible because it is an apparatus claim directed towards a physical circuit. However, the interesting part is in claims 2 and 3, which are both very similar method claims. With claims 2 and 3, the USPTO shows how a method claim can be found invalid by merely claiming the implementation of an abstract idea on a computer, but how a very similar claim can be found eligible by adding a “practical application” of the abstract idea. Here, the “practical application” was the added claim limitation that required using a neural network to take additional steps of performing network packet filtering based on the output of the AI algorithm.

Potential Impact on 103 Analysis: A Look Ahead

I also discussed my thoughts on how AI might change the 103 analysis. To understand how AI might influence the 103 analysis in the future, let’s first review the basics of Section 103:

35 U.S.C. 103 states that a patent may not be obtained “if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains.”

Key elements here are:

• Prior Art: Existing knowledge and inventions in the field.
• Person Having Ordinary Skill in the Art (PHOSITA): A hypothetical person with typical skills and knowledge in the relevant field.
• Obviousness: Whether the invention would have been obvious to the PHOSITA based on the prior art or combinations of the prior art.

With this in mind, here are my thoughts on how AI might impact the 103 analysis:

1. Expanded Access to Prior Art: An AI trained on enough data is going to know a lot… and will be able to understand whether prior art is relevant to a new invention. While this doesn’t change the law, it does effectively raise the bar for proving non-obviousness, as AI systems could potentially uncover more relevant prior art than traditional search methods. In other words, my take is that increased access to prior art is going to make it so that it will be harder to get through prosecution (and hopefully leading to higher quality patents being issued). However, it’s important to note that the mere existence of more prior art doesn’t necessarily make an invention obvious – it’s about how that art would be combined or modified by a PHOSITA.
2. Evolving Definition of PHOSITA: As AI becomes more integrated into various fields, the baseline for “ordinary skill” could significantly increase. This could mean that a PHOSITA might be expected to have access to and be capable of using AI tools in their field, potentially raising the bar for non-obviousness. In other words, if the PHOSITA is charged with the ability to effectively use AI, it becomes easier and easier for the PHOSITA to come up with new inventions.
3. Broadening of Analogous Art: When combining various prior art, the law does not simply allow highly dissimilar inventions to be combined. Instead, a second prior art disclosure must be “reasonably pertinent” to be combined. AI’s ability to draw connections across diverse fields might expand what the law considers “reasonably pertinent” in the obviousness analysis. This could lead to a broader interpretation of analogous art, potentially making more combinations obvious. However, the legal standard for analogous art still requires human judgment about what a PHOSITA would reasonably consider.

What’s Next?

The intersection of AI and patent law is a rapidly evolving space. As AI continues to advance, we can expect further changes in how we approach patentability, especially in terms of inventorship, obviousness, and the scope of prior art. These developments raise fascinating questions about the nature of innovation and the role of human creativity in an AI-augmented world.