Generative AI systems are increasingly accessible to companies of all sizes. Presently, the U.S. Patent and Trademark Office’s (USPTO) guidance notes that AI systems “remain tools [like laboratory equipment, software, databases that are] used by the human inventor who conceived the claimed invention.”[1] When using generative AI in the invention-development process, there are several critical points to remember.

1. Protect your ideas before entering them into an AI system

Inputting invention details into an AI prompt may constitute a public disclosure, which can jeopardize your patent rights, especially internationally. To mitigate this risk, we recommend:

• Use AI systems that keeps your inputs confidential. This way, you are not only avoiding a potential public disclosure to preserve your potential patent rights, but also keeping your ideas confidential.
• Consider filing for a provisional patent application prior to inputting the information into AI, especially if you do not have access to a confidential AI system. This can preserve your potential patent rights for the information within the provisional application.

2. Document human contributions to the invention

An AI system cannot be named as an inventor on a U.S. patent application. Under current USPTO guidance, an inventor is a person who “contributes to the conception of the invention.”[2] The USPTO’s guidance emphasizes that establishing inventorship has always been fact intensive.[3]

To help document human inventorship, we recommend:

• Maintain internal records that make clear an actual human (the inventor) contributed to the conception of the invention.
• Consider whether to file a provisional patent application before utilizing AI for further development, even if it’s a confidential system. Doing so may establish an earlier priority date, prevent inadvertent disclosures as outlined above, and also help establish human inventorship prior to involving AI.

3. Check AI-generated information carefully

AI systems can product extensive output, including inaccurate information or fabricated references. Keep the following in mind when reviewing AI-produced outputs:

• Verify factual claims and cited sources against reliable materials, as AI is known to hallucinate.
• Keep the USPTO’s duty to disclose in mind.[4] If an AI system identifies potential prior art, you have an obligation to disclose that to the USPTO if it is material to patentability, even if discovered after the patent application filing. Note that this can increase government fees if significant amounts of prior art are submitted.

It is important to remember that the above is general guidance about U.S. patent considerations and that the law governing AI use is ever-changing. For help developing an AI-use protocol that protects your organization’s intellectual property, give Tucker Ellis a call.

[1] USPTO Revised Inventorship Guidance for AI-Assisted Inventions; FR Doc. 2025-21457; November 26, 2025 (See, https://public-inspection.federalregister.gov/2025-21457.pdf).

[2] MPEP 2109(II).

[3] FR Doc. 2025-21457.

[4] 37 CFR 1.56