The rapid integration of artificial intelligence into the innovation process has introduced a host of novel challenges for patent law. From questions about who qualifies as an inventor when AI plays a central role, to the rising bar for non-obviousness, in-house counsel must navigate a legal landscape that is evolving in real time. This article examines four critical areas — inventorship, ownership, subject matter eligibility and obviousness — and concludes with practical tips for building a defensible IP strategy in the age of AI.

Inventorship: The Human Requirement

The United States Patent and Trademark Office (USPTO) and U.S. courts have made it unambiguously clear that AI cannot be listed as an inventor on a patent. In Thaler v. Vidal, 43 F.4th 1207 (Fed. Cir. 2022), the Federal Circuit held that only natural persons may be named as inventors under the Patent Act. The USPTO reinforced this principle through subsequent guidance issued in 2024 and 2025. 

Based on the well-known Pannu factors (Pannu v. Iolab Corp., 155 F.3d 1344 (Fed. Cir. 1998), the February, 2024 USPTO guidance confirmed that the use of AI does not negate inventorship, provided a human made a "significant contribution" to the claimed invention. The November, 2025 guidance rescinded the 2024 guidance, and further clarified that the Pannu factors were no longer the standard. Instead, conception must be by a natural person who possesses a "definite and permanent idea of the complete and operative invention."  In other words, a person of ordinary skill in the art (POSA) would have enough information from the inventor to reduce the invention to practice without extensive research or experimentation.  As AI capabilities increase, particularly with the emergence of agentic AI and systems capable of greater autonomy and less human contribution, patent challengers may initially argue that issued patents are legally invalid due to the lack of a natural person who possesses a definite and permanent idea of the complete and operative invention.

Ownership and Improper Inventorship

Inventorship and ownership are inextricably linked. Under both pre-AIA and post-AIA law, improper inventorship can have serious consequences. Under pre-AIA law, improper inventorship could be raised as a ground of invalidity under 35 U.S.C. § 102(f), requiring clear and convincing proof that an unnamed inventor was in fact a co-inventor. While correction was available under 35 U.S.C. § 256, it originally required that the error arose "without any deceptive intention."

Under the AIA, improper inventorship implicates entitlement under 35 U.S.C. § 101 and raises questions of ownership and standing. Correction under § 256 no longer requires a showing of lack of deceptive intent, but the underlying requirement that only true inventors be named remains critical. Improper inventorship remains a basis for invalidity challenges and ownership disputes during enforcement, making accurate identification of human patent inventors a front-line concern for in-house teams. 

The current federal position on AI-generated output is clear: fully autonomous AI-generated output receives no patent protection absent human conception, and listing AI as an inventor in most countries renders a patent invalid. Conversely, where a human uses AI with meaningful creative or inventive control, the resulting innovation is potentially patentable. Human testing, validation, and modification of AI outputs strengthen the inventorship position, while pure prompting alone is usually insufficient. 

Patent-Eligible Subject Matter Under § 101

The question of whether AI-related inventions constitute patent-eligible subject matter under 35 U.S.C. § 101 is developing along divergent lines between the USPTO and the Federal Circuit. The USPTO has taken what might be described as a minimal barrier approach to patenting AI-related claims. In Ex Parte Desjardins, Appeal No. 2024‑000567 (PTAB Sept. 26, 2025, Appeals Review Panel Decision) (precedential) (Nov. 4, 2025), the Patent Trial and Appeal Board (PTAB) held that claims directed to improvements in training a machine learning model comply with § 101 and should be assessed under §§§ 102, 103, and 112. USPTO Director John Squires has publicly supported an expansive approach to eligibility before Congress. 

The Federal Circuit, however, has taken a more cautious stance. In Rensselaer Polytechnic Institute v. Amazon.com, Inc., 2026 WL 506661 (Fed. Cir. Feb. 24, 2026), the court stated that the "generic use of AI without other parameters, such as 'improving the mathematical algorithm or making the machine learning better,' is abstract."

In-house counsel should be aware that the tension between the USPTO's expansive approach during prosecution and the Federal Circuit's more restrictive approach during litigation creates a gap for patentees, such that patents may be issued during prosecution only to face subject matter eligibility challenges later in enforcement proceedings.

Obviousness in the AI Era

AI is also reshaping the obviousness inquiry under 35 U.S.C. § 103. As machine learning tools become widely available and commonly utilized, patent examiners and courts may take the position that AI-assisted optimizations would have been obvious to a skilled practitioner using standard AI tools. Thus, the bar for demonstrating non-obviousness during prosecution may rise, particularly in technology-adjacent fields. In this respect, AI is changing what may count as "routine" innovation, potentially broadening the scope of what is considered obvious.

From a practical standpoint, specific documentation during R&D stages will likely be required to overcome any such obviousness rejections. Applicants should expect increased scrutiny of technical details and enablement in patent applications. Disclosure and claims that were once sufficient to describe software or computational inventions may need to be rewritten to more precisely describe the underlying AI architecture, training methodology, and data prompts and inputs.

Patent drafters face the additional challenge of describing machine learning models and training data in ways that satisfy patent disclosure requirements without giving away proprietary details that the corporate owner would prefer to keep as trade secrets. Strategic claiming, such as by using a mix of system, method, and data structure claims, can help build layered protections.

Practical Tips for In-House Counsel

The following guidance is designed to help in-house counsel build robust, defensible patent programs in an era of increasing AI integration as the law that governs AI continues to evolve.

Tip 1: Establish Internal AI Use Policies. Define how employees and outside counsel are permitted to use AI tools in product development, branding, creative work, and client-facing outputs, and require human review at all critical steps. This is the foundation upon which all other protections are built.

Tip 2: Document Human Contribution Rigorously. Create audit trails that record the nature and extent of human involvement and intentional or unique decision-making in AI-assisted inventions. Documentation should demonstrate that human contributors brought internal company information or particular technical expertise that AI would not have known or considered. Failure to document human contribution increases ownership and validity risks. 

Tip 3: Map AI-Reliant Innovation to Human Conception. Even where innovation is heavily reliant on AI, it must still be mapped to human conception. Practitioners should be ready for higher scrutiny of who conceived what, and patent claims must be aligned with the documented human contribution.

Tip 4: Avoid Red Flags in Patent Applications. Do not include statements in patent applications indicating that AI generated the invention, identified the solution, or autonomously selected the subject matter or claims. Avoid broad or numerous embodiments with no explanation of human selection, and ensure there is no mismatch between claim scope and the human role.

Tip 5: Review and Update Vendor Contracts. Scrutinize data ownership, IP assignment, indemnification, and confidentiality terms in every agreement involving AI tools or services. The patent procurement process should account for the AI tools used at every stage of R & D.

Tip 6: Reevaluate Employee and Contractor IP Assignment Agreements. Ensure that assignment language captures inventions made with the assistance of AI tools and clearly allocates rights in AI-assisted outputs to the company or organization. Traditional assignment language may not adequately address scenarios where AI tools contribute to the inventive process.

Tip 7: Coordinate Across Corporate Functions. Ensure that IP, employment, privacy, and compliance teams share a common understanding of AI-related risk and work collaboratively from consistent policies to effectuate a comprehensive company-wide AI usage framework. Siloed approaches create IP protection gaps that adversaries can exploit.

Tip 8: Maintain Consistency Between Named Inventors and the Record. Ensure that the same natural person requirement is applied consistently to utility, plant, and design patents, and that named inventors are consistent with the conception documented in the prosecution record. If such inventions exist, be cautious about patent priority chains involving foreign filings that may have named only an AI system as inventor. 

Tip 9: Consider Alternative Forms of Protection. Where patent protection is uncertain due to the level of AI involvement, consider trade secret protection, copyright (where human creative contribution is sufficient), or other strategies to protect valuable innovations. A layered approach that combines patents, trade secrets, and contractual protections offers the most resilient coverage and back-up options should the law change to render some current protections ineffective.

Tip 10: Monitor the Legal Landscape Continuously. USPTO guidance, Copyright Office decisions, and case law in the area of AI are evolving rapidly and are highly nuanced. Build in regular review cycles to ensure your policies and procedures comply with the most recent laws. Anticipate future litigation, USPTO rules, and regulatory changes, as the legal frameworks are still catching up with the technology, such that the opportunity to build a comprehensive IP strategy before challenges or disputes arise is now.

Conclusion

The rise of artificial intelligence does not render intellectual property law obsolete, but it makes navigating that law substantially more complex. Human contribution remains legally central across every U.S. IP doctrine and applies to all patent types. Companies that will weather this transition most successfully are those that treat IP governance as a strategic priority: building documentation practices, updating contracts, educating their teams, and working proactively with outside counsel to adapt their current IP policies and programs to a world in which AI is a constant collaborator. Documentation of research and development activity, contract provisions, and governance policies are the three most powerful risk controls available to companies navigating AI and patent laws that will survive the evolving legal landscape over the next three to five years.