In the past year, news coverage on generative artificial intelligence has dominated headlines. Generative AI offers dominant functionality and opportunities to leverage incredibly powerful algorithms to make new content.

One area that business leaders and research and development teams may think to use generative AI is in the realm of new inventions. However, while generative AI offers great potential, it also comes with risk.

Below, we discuss patent and other IP considerations related to using AI for R&D.

Companies are already working with AI in a variety of contexts, including to improve internal processes, develop new methods of internal production, innovate or design elements of contemplated new products and invent new products and services — or improve existing products and services.

But leveraging generative AI creates patent and other IP risks, including:

  • Unintentional release of company proprietary information;
  • Infringement of existing third-party patent or other IP rights;
  • Loss of ownership of company IP;
  • Invalidity challenges to company IP; and
  • Inappropriate use of AI by employees.

To understand these issues, we first examine how generative AI functions.

The Black Box Problem

Generative AI leverages machine learning technology to collect, process and analyze large amounts of data. The AI then generates content based on said analysis and in response to user prompts.

For example, a user prompt for a chicken parmesan recipe will initiate deep learning models to process and analyze the language prompt and generate output based on learning from underlying training datasets. Generative AI relies on neural networks to spot patterns within data and make predictions based on the pattern analysis.

However, even the developers of generative AI tools often have little insight into the way a generative AI engine makes a given decision — i.e., how it absorbs data, learns from it and generates requested content in response to prompts.

The average user of generative AI tools does not know how outputs are obtained or which data was leveraged to create the output.

This creates risks for generative AI users leveraging the technology to create IP. Generative AI tools need enormous amounts of data to perform effectively.

The most popular generative AI tools leverage content — text or images — obtained from the internet (copyrighted or not); user-generated content, such as social media posts or reviews; or code sets available online — most of which are subject to open source licenses.

As further discussed below, leveraging generative AI can cause patent and other IP risks.

The Patent Landscape

Recent USPTO Guidance, Inventorship and Disclosure Requirements

It is well settled that an inventor of a patent in the U.S. must be a human individual.[1] While this does not necessarily prevent the use of AI’s assistance in creating inventions outright, it does prevent companies from seeking patent protection on inventions solely created by an AI bot.

On Feb. 12, the U.S. Patent and Trademark Office released the following inventorship guidance for AI-assisted inventions, explaining that “while AI-assisted inventions are not categorically unpatentable, the inventorship analysis should focus on human contributions, as patents function to incentivize and reward human ingenuity.”

According to the USPTO, a natural person must have significantly contributed to the conception of each claim in a patent — for patents with a single inventor who has used an AI system to create the invention, the human inventor must have “significantly contributed” to the conception of each claim.

To determine whether an inventor’s contribution was significant, applicants are guided to use the factors under the U.S. Court of Appeals for the Federal Circuit‘s 1998 Pannu v. Iolab Corp. decision[2] and a nonexhaustive list of principles to aid in the application of the Pannu factors to the AI-assisted invention context.

For example, “simply owning or overseeing an AI system” used in the context of inventing, without more, does not make that person an inventor, according to the USPTO guidance.

However, a significant contribution could potentially be shown by creatively constructing a prompt because of a specific problem.

Notably, the USPTO guidance contemplates a situation in which a person designs, builds or trains an AI model to solve a specific problem as potentially rising to the level of a significant contribution.

The USPTO has a variety of strict disclosure requirements when seeking a patent. Applicants are required to disclose details about themselves, any inventors and any public information that may be materially related to the invention they are pursuing.

The USPTO has not changed its disclosure requirements, but the guidance reminds applicants of the existing duty of disclosure in the context of inventorship.

For AI-assisted inventions, parties with a duty to disclose should make sure they are properly assessing inventorship in view of the guidance.

Obviousness Issues

The use of generative AI also raises questions about what makes an invention obvious. Under Title 35 of the U.S. Code, Section 103, the USPTO will reject patents that are obvious — meaning that it would be obvious to a person of ordinary skill in the relevant field to combine different ideas or components to make the invention.

The U.S. patent system rewards so-called inventive sparks — as opposed to merely applying an obvious solution to a different context.

It could be argued that ideas generated by AI are obvious. Arguably, AI bots are not creative; rather, they analyze large swaths of existing information and use an algorithm to make a prediction.

They do this without any specialized knowledge in a given field. Some may argue that AI-assisted inventions are therefore obvious. While courts have not weighed in on this issue, AI-assisted inventions may eventually be seen as obvious due to the prior art analyzed by the AI and the automatic process or algorithm combining such art.

Infringing Other IP

In addition to the risks outlined above, companies must also consider third-party IP rights when leveraging generative AI. As described above, generative AI leverages models trained on large amounts of information and data from a variety of sources, which can include third-party IP.

Even AI developers are unable to assess where all of the training data originates and, therefore, cannot guarantee that the outputs are free of infringement risk.

Companies looking to implement AI using third-party vendors should explore contractual or technical safeguards to help mitigate this risk.

Loss of Confidential Information

If the risks on the patent side were not enough, generative AI also presents a risk to the proprietary and confidential nature of information if not used properly.

If an AI bot is given unfettered access to an internal network or process, it may use that information while generating other content. AI bots seek to learn from as many sources as they can and will leverage any information provided, even if the next ask is coming from a third party or a competitor.

These types of risks are already playing out in the real world. Many companies have already banned or restricted the use of generative AI due to potential or actual leaks of company confidential information. Many AI vendors offer enterprise solutions that may sufficiently lower this risk.


In the 2022 Thaler v. Vidal decision, the U.S. Court of Appeals for the Federal Circuit ruled that an inventor of a patent in the U.S. must be a human individual — so, no robots allowed.

While this does not necessarily prevent the use of AI’s assistance in creating inventions outright, it does prevent companies from seeking patent protection on works invented by an AI bot.

To the extent you are leveraging AI in your business to create new proprietary products, it’s important to ensure that the AI is used as an assistive tool and that inventors are not claiming concepts generated by AI in patents.

Conclusion: Practice Tips When Using AI

To conclude, after considering all of the above, a potential AI user may have second thoughts about using AI. However, generative AI can still be a powerful tool and risks can be mitigated if used correctly. When using AI, companies should consider the following six practice tips in order to protect their IP rights and confidential information:

1. If using AI to create or within R&D processes, ensure it is used as a supplementary tool and is not contributing to invention conception.

2. Implement technical safeguards to ensure controlled and intentional access by AI with respect to company systems and information.

3. Implement contractual protections and safeguards with AI vendors to minimize the risk of disclosure and use of company confidential information.

4. Educate employees about appropriate AI prompts as well as information that should not be used.

5. Make sure to document your use of AI in the event it is needed in a legal context in the future.

6. Stay up-to-date on evolving AI-centered regulations.

Mackenzie Martin is a partner at Baker McKenzie LLP and co-leads the firm’s global patent practice.

Bryce Bailey is an associate at the firm.

The opinions expressed are those of the author(s) and do not necessarily reflect the views of their employer, its clients, or Portfolio Media Inc., or any of its or their respective affiliates. This article is for general information purposes and is not intended to be and should not be taken as legal advice.

[1] Thaler v. Vidal , 43 F.4th 1207 (Fed. Cir. 2022), cert. denied, 143 S. Ct. 1783 (2023).

[2] Pannu v. Iolab Corp. , 155 F.3d 1344, 1351 (Fed. Cir. 1998).

*Article originally published by Law360: Averting Patent And Other IP Risks In Generative AI Use – Law360*


Mackenzie Martin co-leads the Firm's Global Patent Practice and is a member of the Firm's North American Intellectual Property & Technology and North American Trial Team Steering Committees. She focuses on intellectual property litigation, counseling, licensing, and portfolio strategy matters. She has significant experience in patent and trade secret litigation actions in federal district courts, before the US International Trade Commission, and before the United States Patent and Trademark Office Patent Trial and Appeal Board.


Bryce joined Baker McKenzie's IP Practice Group following his graduation from the University of Georgia School of Law in 2020. With an electrical engineering degree, practical work experience for various electrical contractors, and studies in circuitry design and electromagnetics, he is able to leverage his experience across the patent prosecution, IP litigation, and technology transactions areas.