Co-Creation and the Patent Process

In most countries, and under current U.S. law, only natural persons can be named as inventors on a patent. AI systems cannot be listed as inventors, even if they autonomously generate inventions.

Patent protection is available for AI-assisted inventions if a human made a “significant contribution” to the inventive concept. Under U.S. law, an inventor is one who contributes to the conception of the invention, i.e., made a significant contribution to the conception of the invention. Reduction to practice — the process of demonstrating that an invention works for its intended purpose — per se, is irrelevant, except in instances where conception and reduction to practice occur simultaneously. As such, conception is the touchstone of inventorship, and simply creating the prompt that leads AI to generate what becomes an invention likely does not rise to the level of making a significant contribution to that invention’s conception.

The Pannu Factors

To show that a natural person has made a significant contribution to an invention that qualifies them as a co-inventor, this person must satisfy the three Pannu Factors:

1. Played a meaningful role in the creation of the invention.
2. Contributed to the problem-solving process.
3. Was involved in the reduction to practice.

A human co-inventing with AI cannot simply prompt the AI system with a problem to solve. Rather, the human inventor’s involvement should, for example, provide specific constraints on solutions that the AI system can suggest, iterate on the solution with a series of prompts (preferably detailed prompts), or further refine the AI system’s output in a significant way.

As with all inventions, the requirements for patentability still stand: novelty, non-obviousness, and utility. And the subject matter itself must be patentable — for instance, abstract ideas such as algorithms are not patentable subject matter, whether they are human or machine-assisted creations.

A Legal Landscape in Flux

It should, however, be noted that this is a quickly evolving area of the law. As with any technology, the law follows the technology. Countries around the world are wrestling with how to handle these co-inventions. For example, officials in India are currently debating and discussing the potential for recognizing AI as an inventor. In the U.S., the current United States Patent and Trademark Office (USPTO) leadership has indicated it may be planning to give AI-“created” inventions greater scrutiny and, potentially, more leeway in the overall patent regime.

Although there are few if any countries that will award patents to solely AI-created inventions, the field is evolving. Currently, the U.S., European Patent Office, India, the U.K., Japan, and Australia do not recognize AI as a sole inventor. China, likewise, does not recognize AI as a sole inventor, but that country is developing regulations and guidelines that seem to be coming closer to recognition. Nevertheless, the Chinese Patent Laws and China National Intellectual Property Administration guidelines continue to require that civil subjects (natural persons) be inventors for purposes of obtaining patent protection.

Basic Strategic Considerations

Because AI cannot be identified as an inventor in most, if not all, jurisdictions, IP ownership continues to fall under the laws (whether specific patent laws or general civil/statutory laws) on a country-by-country basis. As such, inventions may be owned by an inventor, an inventor’s employer, or another entity to which an invention and/or patent application may be assigned. Thus, ownership must still be evaluated based on traditional ownership principles.

With this in mind, best practices dictate that human creators should document in detail how they are using AI in creating content, and not just patent-related content. These detailed records should:

• Capture how humans contributed to the invention process by including notes, design iterations, and other evidence of how human ingenuity played a role in decision-making at every step.
• Document the prompts used with AI to show that human direction was involved in every iterative step, and that a human made decisions about course and direction changes.
• Provide granular detail on the specific work done by each human involved at every step. 

It is important to avoid making claims for an invention generated solely by AI, particularly in the case of independent claims. If you determine that an AI-assisted invention cannot be patented, consider maintaining the invention as a trade secret. And if the invention is protected by trade secret, be aware that AI platforms may be trained on publicly available data — which means your trade secret may potentially be vulnerable.

Using AI to Draft Patents and Analyze Patent Specifications and Claims

Numerous AI tools currently available can be used to draft patent applications and analyze claims, and more tools become available every day. Here are some important things to know about relying on these tools:

• Ethical duties still apply. While there is currently no duty to inform the USPTO that an AI tool was used to draft a document, ethical duties still apply, including the duty of candor and good faith. No matter how sophisticated and advanced they become, these tools still require human judgement, input, attention, and review.
• AI tools provide powerful assistance with analysis. These tools can conduct freedom to operate (or right to use) evaluations, provide claim-by-claim analysis on multiple patents, identify relevant prior art references, map claims, provide foreign language translations, and perform other analysis tasks.
• AI tools can also help with strategy. These tools can conceptualize and carry out search strategies and identify “white spaces” in which patent protection may be sparse and innovation may be open for study. AI can also be used for competitor analysis, monitoring competitors’ portfolio development, and other strategic steps.
• But AI also makes mistakes. Alongside enjoying the benefits of these tools, practitioners must be mindful of the risks associated with AI, such as “hallucinations” and errors, and practitioners are responsible for correcting inaccuracies. Simply relying on an AI’s output is not sufficient to meet the duty of reasonable inquiry under USPTO rules.
• AI tools do not protect confidential information. Practitioners must protect client information, and using public AI tools may lead to inadvertent disclosure of confidential materials, trade secrets, and the like.
• AI tools sometimes violate copyright. AI platforms may not “recognize” when output includes third-party copyright-protected materials that go beyond fair use. In some cases, wholesale copying of texts and other materials have occurred, and lawsuits have followed.
• The practitioner must still sign off. All AI-generated materials must be reviewed and signed by a qualified individual before filing, so be sure you have evaluated the material carefully and are comfortable putting your name on it.

Some jurisdictions require disclosure of the use of AI in patent drafting, and it is a good practice to follow regardless of where you are based. Such up-front disclosures create transparency and prevent a “cloud” over the patent later on, especially during enforcement. Note that in the U.S., misuse of AI in inventing and patent drafting may violate the duty of candor and good faith required when practicing before the USPTO.

Questions about using AI in your content creation or patent analysis process? Reach out to Mitchell Weinstein or another member of LP’s Intellectual Property Group.