USPTO Issues New Inventorship Guidance for AI-Assisted Inventions (November 2025): What It Means for Innovators & Patent Practitioners

The USPTO’s new 2025 inventorship guidance (Fed. Reg. notice 2025-21457) resets the treatment of AI-assisted inventions by rescinding the 2024 AI inventorship guidance and re-anchoring inventorship solely in traditional human conception. This is a significant policy shift with major implications for R&D teams, patent prosecutors, and global portfolio managers.

Below is a breakdown of what changed and how to navigate it.

Key teachings:

1. Only humans can be inventors, no exceptions

 The Office reiterates Thaler v. Vidal. AI systems can assist, but they cannot conceive. Even if an AI autonomously outputs inventive subject matter, no inventorship can attach unless a human forms the definite and permanent idea of each claimed feature.

2. No special or modified test for AI-generated or AI-assisted inventions

The USPTO now applies the traditional conception standard used for any other invention. No carve-outs, no AI-specific inventorship threshold.

This settles confusion created by the 2024 guidance, which attempted to adapt inventorship to AI contexts.

3. Pannu factors are out for AI cases

The 2024 approach tried applying Pannu v. Iolab (a joint-inventorship case between human co-inventors) to AI-human collaborations. The 2025 guidance clarifies:

• Pannu applies only to human co-inventors,
• Never to AI, because AI cannot be a legal co-inventor.

4. AI is now formally treated like any other research tool

The USPTO explicitly analogizes AI to:

• lab equipment
• data-processing tools
• software
• scientific literature

AI may accelerate, model, optimize, propose, or generate, but this does not create inventorship.

5. Applies to all patent types

This includes:

• Utility
• Design
• Plant

No category-specific exceptions.

6. Priority claims can be impacted

If a foreign filing lists only an AI system as the inventor, a U.S. application cannot claim priority to it. A valid priority claim requires at least one common human inventor across filings.

This will matter for portfolio managers working with filings from jurisdictions experimenting with AI inventorship rules (e.g., Australia, South Africa).

Practical Implications for Innovators, R&D, and Patent Practitioners:

1. Document the human’s inventive contribution

Maintain contemporaneous documentation showing:

• What the human conceived
• How the AI was used
• How the human selected, interpreted, or refined AI output
• Why the resulting claim elements reflect human conception

This mirrors record-keeping expectations for complex research environments.

2. Align claims with what the human actually conceived

Draft claims so that every claimed feature corresponds to a human’s cognitive contribution—not to:

• unexplained AI output
• autonomous AI design decisions
• optimized structures never understood by a human inventor

3. Educate engineers and scientists

Clarify internally:

• Using AI tools does not diminish inventorship
• AI-only inventions are not patentable in the U.S.
• Human conception must be demonstrable for each claim element

4. Review foreign filings carefully

Before filing a U.S. application:

• Check who is listed as the inventor internationally
• Ensure at least one human inventor is common
• Correct foreign inventorship where possible (some jurisdictions allow later amendment)

Why This Guidance Matters:

This 2025 update restores predictability for applicants and acknowledges the reality that AI is a ubiquitous R&D tool. It avoids stretching inventorship doctrine into AI-specific exceptions while preserving:

• Legal clarity
• Consistency with longstanding case law
• The constitutional requirement that patents issue only to human inventors

In short: AI may innovate but only humans invent.

Comparison with AIPPI resolutions:

The most relevant resolution related to the USPTO guidance is the resolution to Q 272 “Inventorship of inventions made using Artificial Intelligence” (2020), although another significant resolution for AI inventions was Resolution Q276 “Inventiveness and sufficiency of disclosure in AI inventions” (2021).

There is significant harmony between the USPTO’s 2025 position and the AIPPI Q272 resolution regarding the fundamental nature of inventorship. Both the resolution and the USPTO guidelines establish that AI systems cannot be named as inventors or joint inventors. The USPTO classifies AI as a tool or instrument used by humans, analogous to laboratory equipment, which aligns with AIPPI’s view that an invention should not be excluded from protection merely because AI contributed to its creation. Likewise, by applying the traditional “conception” standard to all inventions regardless of AI use, the USPTO guidance is consistent with the resolution that the requirements for a natural person to be an inventor should not differ based on the use of AI tools.

AIPPI’s Resolution Q272 suggests that “to foster innovation, inventions made using AI should not be excluded from patent protection per se, regardless of whether or not there is sufficient contribution by a natural person to be named as an inventor and provided that there is a natural or a legal person named as an applicant”.  The USPTO guidance does not provide a path for a “legal person” to be the primary applicant for an AI-only invention in the absence of a human inventor.

No standards related to sufficiency of disclosure or inventiveness are discussed in the USPTO guidance, which is understandable considering that it specifically addressed inventorship. This means that there is still room for more guides on the evaluation of the intervention of AI systems and enablement of claimed AI related inventions. However, the concept of the person skilled in the art discussed in AIPPI’s Q 276 as a hypothetical natural person who would use AI technology only if it were “expected from the average person in the field” relates to the USPTO concept of a human inventor’s mental conception and the evaluation on whether it was complete enough that only “ordinary skill” was required for reduction to practice, but it does not formally update the “person skilled in the art” standard for AI-heavy technical fields.