Updates in Artificial Intelligence and IP

Apr 4

Artificial intelligence (AI) is still in its infancy, but growing rapidly. With the rapid change in the capabilities of different AI systems, reactions from the IP community and IP agencies are constantly changing to address the updates. Recent developments in AI have changed IP in analyzing standard essential patents, performing patent searches, and copyright ownership.

Standard Essential Patents (SEP)

Patents in areas that are viewed as essential to our current society, and which serve as the core of industry-standard technologies, are considered Standard Essential Patents (SEPs). These SEPs are subject to certain licensing rights to allow conformance with standards by others in the field. Some prevalent examples of SEPs include USB, Wi-Fi, and 4G/5G technologies. As companies and inventors work towards securing new patents, they often have to consider if their advancements will become a standard or essential in that field and, therefore, whether they have special disclosure rules for their application. Some inventors and patent firms have considered turning to AI to determine if new generative models can determine whether their applications will be considered essential.

Humans who undertake studies to determine essentiality often have a success rate in 4G and 5G technologies reported from 50% to as low as 8%, but this can vary for different technology fields. This low success rate is part of what is pushing inventors to look for other potentially better methods, including AI. Unfortunately, at this time, most AI systems have performed worse than their human counterparts at determining if a patent will be deemed essential.

A large factor in this performance seems to be the relatively low sample size the AI is able to pull from to create a model. Where AI has helped is in the efficiency of the process for a human. AI systems are able to scour numerous patents or other documents for specific technical information that was flagged as making the patent essential. By doing this, the AI can speed up the process the human assessor undertakes in gathering the information needed in a fraction of the time.

As time passes and more samples are created to provide more data points, there is some hope that AI can help determine the essentiality of a patent with even more precision. However, there is also the chance that as new technologies are deemed essential, the technical features of each field are so diverse and divergent that there are never enough samples in a given field for an AI to perform the assessment functionality alone.

“AI ownership is a hot topic in all aspects of IP. Do works created by AI carry the same weight, and therefore protections, as those created by humans? According to a new official policy from the United States Copyright Office, that answer is no.”

Patent Searching

Similar to SEPs, many inventors and companies want to know the patent landscape of their field before they spend the time and money to proceed with a patent. Often in these cases, they will elect to perform a patent search. Patent searches are carried out using tools that simply pull keywords determined by the human searcher to find as many hits in patents and publications. The human searcher then must go through the located documents to determine how relevant the results are to the proposed product.

Attempts to use ChatGPT to perform patent searches which rely solely on a short description of the invention return results that are irrelevant and not helpful in determining whether to proceed with a patent application. Similar to SEPs, ChatGPT can be very helpful in narrowing down and saving time, but it currently is not able to fully perform the function. One of the most important things to note when using AI such as ChatGPT is the AI is not able to discern the technical benefits of an invention from a brief description. The AI system looks only at the listed physical elements but does not infer how they work together, or the benefit of having the included elements. Discerning the technical benefit of an invention is one of the most important tasks in a patent search as, at times, different elements are being used to perform the same benefit.

One of the areas that ChatGPT performs strongly is the extraction of keywords. If ChatGPT is provided the same short description as above, it can quickly pull a list of desired keywords that can then be entered into a patent search. While just finding the keywords is not the most helpful, ChatGPT is also able to develop a search formula including the extracted keywords that would allow a person to perform an effective patent search.

AI systems are not currently in a place to replace humans, but they can assist and create more efficient flows of data. The more efficient the patent search is, the more informed the decision to proceed or not with the patent can be.

Copyright Ownership

AI ownership is a hot topic in all aspects of IP. Do works created by AI carry the same weight, and therefore protections, as those created by humans? According to a new official policy from the United States Copyright Office, that answer is no. Officially released on March 16, 2023, the Copyright Office issued an official policy on AI ownership and specific rules around how AI may be used.

In the policy, the Copyright Office goes into detail that works created solely by a machine are not eligible for protection, but works created by humans with machine assistance may still be eligible for copyright protection. One example used by the Copyright Office was a recent comic book submitted for protection in which all of the images were generated by an AI. In that case, the Copyright Office extended protection to all of the written words, as well as to the layout and flow of the comic book. The images themselves were not allowed to be protected as those were generated by a machine alone.

The Copyright Office noted that applicants now have a duty to disclose the use of any AI tools in the creation of the work. There will now be a portion of the application titled “Author Created” where the author must disclose what portions were human-created, whereas AI-generated content must be excluded in the “Limitation of the Claim” section of the application.

The policy does clarify that this does not exclude certain AI tools that artists and authors can use to assist in the creation of works. Portions of Adobe Photoshop have AI tools to help artists edit images, and the editing of the image would in no way limit the protection offered to the artist.

AI-created art and the protections around it are certainly hot issues in the creative community. The Copyright Office policy seeks to offer some clarity to this issue regarding protection. As AI art continues to evolve the Copyright Office may have to further amend and consider this policy and if it stands to protect artists and authors in the best possible way.