Introduction
For nearly two years, companies building products with generative AI had a specific framework to navigate inventorship decisions: the Biden-era February 2024 USPTO guidance on AI-assisted inventions. It introduced a ‘significant contribution’ analysis — a structured framework for assessing when the involvement of an AI system in the inventive process affected inventorship determination. It wasn’t perfect, but it gave practitioners something to work with.
On November 28, 2025, the USPTO rescinded it. The revised guidance is straightforward: AI systems are tools, not inventors. The traditional conception standard — a natural person must conceive each claimed invention — applies regardless of how much AI was involved in the process. No separate AI-specific analysis. No Pannu factor framework for evaluating AI contributions. The standard that applied before February 2024 applies again now.
For companies with existing or pending patent portfolios that were navigated under the 2024 framework, the reset has specific implications. Some inventorship determinations may need to be revisited. Some pending applications may need claim structure review. And the documentation practices that the 2024 framework encouraged — tracking human vs. AI contributions — are now more important, not less, because the burden of demonstrating human conception falls entirely on the applicant with no framework to assist the analysis. This article covers what changed, what needs to be reviewed, and how to build a patent portfolio strategy that holds up in the post-reset environment.
What the November 2025 USPTO Guidance Actually Says
Let’s be precise about what was rescinded and what wasn’t. The November 2025 guidance rescinds the February 2024 Inventorship Guidance for AI-Assisted Inventions and removes the Pannu factor analysis that had been applied to AI inventorship questions under that framework. It does not change the fundamental inventorship standard. What it changes is the AI-specific overlay that the 2024 guidance had built around it.
February 2024 framework (rescinded): AI-specific ‘significant contribution’ analysis required. Practitioners assessed whether AI contributions were significant enough to affect inventorship. Structured Pannu factor framework provided guidance on how to evaluate human vs. AI contributions to each claim. Created a separate analytical pathway for AI-assisted inventions.
Post-November 2025 standard: No AI-specific analysis. The traditional conception standard applies: a natural person must form a definite and permanent idea of the complete and operative invention in their mind. AI systems are laboratory equipment or research databases — they assist the inventive process but cannot conceive inventions. If the human contributor’s involvement amounts to merely running an AI system and taking its output without meaningful creative selection or modification, that is unlikely to meet the conception standard.
The practical test remains unchanged: did the named human inventor(s) conceive the claimed invention? Conception means the formation in the mind of the inventor of a definite and permanent idea of the complete and operative invention. In an AI-assisted workflow, the question is whether the human inventor’s contribution amounts to genuine conception — selecting, modifying, refining, and directing the AI’s output toward a specific inventive idea — or whether the human merely prompted the AI and accepted what it produced.
That line has always existed. The 2024 framework provided structure for navigating it. The post-reset environment removes that structure and leaves the determination to the traditional standard, applied case by case.
The Documentation Imperative: Why Internal Records Now Carry More Weight
The 2024 framework provided a structured analytical pathway for AI inventorship decisions. Without it, the burden of demonstrating human conception sits entirely on the applicant — supported by internal documentation that the USPTO or a challenger in litigation can scrutinise.
This is not a new requirement. Invention records have always been important. What’s new is the specific evidentiary gap the 2024 rescission creates: companies that relied on the 2024 framework’s structured analysis to support their inventorship determinations are now operating without that framework. The internal records that document human conception — or that should document it — are the only thing that remains.
Here’s what those records need to show for an AI-assisted invention to withstand scrutiny:
- What the AI produced. The specific outputs — text, code, designs, molecular structures, whatever the AI generated — should be documented at the time of generation. Screenshots, log files, prompt-output pairs. The record should show what the AI contributed before human engagement.
- What the human inventor contributed. The record should show what the human inventor selected from the AI output, what they modified, what they discarded, and what creative or technical judgment they applied in directing the AI toward the claimed inventive concept. This is the conception evidence — the documentation that the human formed a definite and permanent idea of the invention, not just ran a prompt and accepted the result.
- The inventive concept that the human conceived. The invention disclosure form should clearly articulate the inventive concept — the specific technical advance that the patent will protect — in terms that reflect the human inventor’s understanding and contribution, not just a description of the AI’s output.
- The connection between AI output and human conception. The chain from what the AI produced to what the human inventor conceived and claimed should be traceable in the internal record. If an examiner or a litigation opponent asks ‘how did the human inventor conceive claim 1?’ the documentation should provide a clear, specific answer.
Portfolio Audit: Reviewing Patents Filed Under the Biden-Era Framework
Companies that have existing patents or pending applications that were filed, prosecuted, or inventorship-determined under the 2024 framework need to assess those assets under the post-reset standard. The rescission doesn’t automatically invalidate existing patents. But it does create a different lens for evaluating inventorship correctness.
For patent portfolio analysis purposes, the review should cover three categories:
- Patents granted under the 2024 framework where AI played a significant role. Identify patents in your portfolio where (a) AI tools were extensively used in the inventive process and (b) the inventorship determination relied on the 2024 framework’s structured analysis. Assess whether the named human inventors meet the traditional conception standard without the 2024 framework’s support. Where inventorship appears correct under the traditional standard, document why. Where it may not be, consider whether correction is appropriate and what the risk profile of leaving it uncorrected is.
- Pending applications in prosecution where inventorship may need revisiting. Applications that are still pending — not yet granted — can be corrected before grant more straightforwardly than granted patents. Identify pending applications where the inventorship determination relied on the 2024 framework and assess whether the traditional standard is met. The pre-grant window is the lower-risk opportunity to correct if needed.
- Applications where AI contribution was overstated or understated. The 2024 framework created two-directional risk: some companies may have attributed inventorship to human contributors who didn’t meet the conception standard (overstated human contribution), while others may have excluded human contributors whose contribution genuinely met the standard but was framed as ‘AI assistance’ (understated human contribution). Both directions of error need to be assessed.
“The rescission of the 2024 guidance doesn’t change what the right inventorship answer is — it changes the framework that was being used to get there. Companies that arrived at correct inventorship determinations under the 2024 framework are fine. Companies that relied on the 2024 framework’s structure to support determinations that don’t hold up under the traditional conception standard have a problem.”
Filing Strategy: Building a Portfolio That Holds Up Post-Reset
Going forward, AI-assisted patent portfolio strategy needs to be built around the traditional conception standard — without the 2024 framework’s structural support. Here are the five adjustments that matter most for drafting patent applications for AI-assisted innovations under the post-reset environment.
1. Structure claims around human-conceived inventive concepts — Claims should be directed to the specific inventive contribution of the human inventor — the technical advance they conceived, directed, or selected from AI-assisted exploration — not the AI’s output in general. Claims structured around specific human inventive contributions are more defensible under the conception standard than claims that describe the AI’s output without a clear line to human inventive decision-making.
2. Document the inventive contribution chain for every AI-assisted application — Before filing, create an internal record showing the path from AI assistance to human conception: what the AI produced, what the human inventor selected or modified, what creative or technical judgment the human applied, and what the resulting inventive concept is. This documentation is the evidence base for inventorship decisions and for defending those decisions in prosecution or litigation.
3. Consider trade secret protection for AI-generated outputs that don’t meet the conception standard — If the commercial differentiator is an AI-generated output that doesn’t represent meaningful human conception — a model architecture, a discovered compound, a generated design — trade secret protection may be more appropriate than patent protection. Patent disclosure requires sufficient human conception to support inventorship. Where that standard isn’t met, filing anyway creates invalidity risk.
4. Review inventorship on pending applications before grant — Pre-grant inventorship correction is significantly simpler than post-grant correction. For pending applications where inventorship was determined under the 2024 framework, assess under the traditional standard before grant and correct if needed. The correction cost is low before grant. After grant, it requires a formal proceeding.
5. File continuation applications structured around clearly human-conceived claims — For existing patents covering AI-assisted technology where the original claims were structured under the 2024 framework, continuation applications can be filed with claims more clearly directed to human-conceived inventive concepts. This builds a stronger portfolio foundation for the post-reset environment while preserving the priority date of the original filing.
The Broader Context: How Other Jurisdictions Handle AI Inventorship
The USPTO’s post-reset position — AI as tool, human conception required — is consistent with the approach taken by every major patent jurisdiction. Understanding each one matters for companies building global AI-assisted patent portfolios.
- EPO: Technical character required, AI cannot be inventor. The EPO requires that patents cover a technical invention — AI-related applications must demonstrate a technical character beyond pure mathematical methods. The EPO has consistently held that only natural persons can be inventors (confirmed in the Dabus applications). European AI patent strategy needs to demonstrate a specific technical effect and ensure human inventors are properly identified.
- China CNIPA: Natural persons only — now explicitly required. As covered earlier in this series, the January 2026 CNIPA guidelines explicitly ban AI systems, entities, or groups from being listed as inventors, and require natural persons with verifiable identity information. China’s position aligns with the post-reset USPTO standard but goes further in requiring explicit identity verification.
- UK: Dabus cases and their aftermath. The UK Supreme Court confirmed in Thaler v. Comptroller-General that AI systems cannot be inventors under the Patents Act 1977. The inventor must be a natural person. AI-generated inventions with no human inventive contribution are not patentable in the UK under current law.
- Australia: Recent alignment with international consensus. The Full Federal Court of Australia reversed the earlier decision that allowed an AI to be listed as inventor, bringing Australia into alignment with the international consensus: only natural persons can be inventors.
The global picture is consistent: no major patent jurisdiction allows AI inventorship. Companies building global AI-assisted patent portfolios need to ensure that human conception is documented and defensible in every jurisdiction they file in — not just the US.
How Our Portfolio Analysis Service Supports the Post-Reset Environment
Our patent portfolio analysis service covers AI-assisted patent portfolio reviews in light of the November 2025 USPTO guidance reset — identifying patents and applications where inventorship may need to be revisited, assessing claim structure against the traditional conception standard, and recommending continuation or correction strategies where appropriate.
For technology companies and startups building AI-intensive product portfolios, we structure portfolio analysis engagements around the specific R&D workflows where generative AI is integrated — mapping the human inventive contribution at each stage of the workflow against the traditional conception standard. The goal is to identify where the portfolio is solid, where it carries inventorship risk, and where filing strategy needs to be adjusted to build on the post-reset foundation rather than the rescinded 2024 framework.
For companies with pending applications still in prosecution, we provide pre-grant inventorship review — the lower-cost window to identify and correct inventorship before grant locks in a potentially vulnerable determination.
Building a patent portfolio with generative AI in your R&D workflow? The November 2025 USPTO reset means the 2024 framework you may have been relying on no longer applies. Our portfolio analysis service maps your AI-assisted inventions against the traditional conception standard — and identifies what needs to change. → Contact Us
Conclusion: The Takeaway
The November 2025 USPTO guidance reset is not a minor administrative update. It removes a framework that technology companies and their patent counsel had been using for nearly two years to navigate inventorship decisions for AI-assisted inventions. In the post-reset environment, the traditional conception standard applies — without structural support, without a separate AI analysis pathway, and with the full burden of demonstrating human conception sitting on the applicant’s internal documentation.
For companies that navigated inventorship correctly under the 2024 framework — that documented genuine human conception and applied the standard carefully — the reset changes little in practice. For companies that relied on the 2024 framework’s structure to support determinations that don’t hold up under the traditional standard, the reset creates inventorship risk they may not have identified yet. The portfolio audit is the starting point for knowing which category you’re in.
AI is a powerful tool for innovation. The patent system still requires a human inventor. Making sure those two realities align in your portfolio documentation is the work that needs to happen now.
