AI-assisted R&D
AI-Assisted Inventorship After the 2025 USPTO Guidance
Using AI does not create a separate inventorship test. The practical question is still which natural persons contributed to the conception of the claimed invention.
By Stewart Myers, Oregon attorney and registered patent attorney · Originally published August 2025 · Last reviewed
Short answer: AI cannot be named as an inventor. Only natural persons may be inventors. The ordinary inventorship framework applies whether or not an AI system helped a team research, model, draft, analyze, or test an idea.
What changed in November 2025?
The USPTO rescinded its February 2024 AI-assisted-inventorship guidance in full and replaced it with revised guidance. The revised approach does not use a special AI-specific “significant contribution” checklist. It applies the same inventorship law used in other patent matters, including the ordinary conception inquiry and the requirement that a named inventor be a natural person.
That means AI use is not, by itself, an inventorship defect. It also means a company should not assume that the person who selected, configured, trained, licensed, or supplied an AI tool automatically becomes an inventor. Inventorship turns on contribution to the conception of what is actually claimed, based on the particular facts.
The analysis is claim-specific and person-specific
A useful review begins with the people and the claims—not the software label. Different patent claims can have different inventors. A person may contribute to one claimed combination but not another, and routine execution, supervision, data entry, testing, or recognition of a result may or may not amount to conception depending on what the person actually contributed.
Prompts, model outputs, lab notebooks, code history, design reviews, experimental records, and invention disclosures can be relevant evidence. None is a universal shortcut. A prompt is not automatically conception, and an output is not automatically proof that no human conceived the claimed invention.
Five practical steps for an R&D team
- Record the human problem framing. Identify who defined the technical problem, constraints, objectives, and proposed solution architecture.
- Preserve meaningful iterations. Keep an appropriate record of who evaluated outputs, rejected alternatives, changed the approach, and supplied technical features that later matter to a claim.
- Separate tool operation from inventive contribution. Document what a person contributed to the claimed concept, not merely that the person used, bought, trained, or maintained a tool.
- Revisit inventorship when claims change. A filing draft, continuation, amendment, or newly emphasized embodiment can change the relevant contribution analysis.
- Coordinate ownership and federal-funding records. Inventorship, assignment, employment obligations, award scope, reporting, and patent strategy are related but distinct questions.
For federally funded R&D, connect more than inventorship
If AI-assisted work occurred under or alongside a federal award, the team may need to map funding periods, statements of work, contributors, inventions, patent families, disclosure records, assignments, and commercialization activity. An award does not automatically make every invention a subject invention, and AI use does not automatically create a federal reporting problem. The funding agreement, agency instructions, current law, and facts control.
When a focused review can help
A fact-specific review may be useful before filing, during claim development, when correcting an invention disclosure, in diligence, or when contribution records are fragmented across founders, employees, contractors, universities, collaborators, and AI systems. Depending on the issue, patent counsel, employment or contract counsel, federal-grants specialists, or other advisers may need to coordinate.
Official source
Focused, not broad AI law
Review the human-contribution record in context.
Cognivito's AI-Assisted Inventorship module is a separately scoped part of Federal Innovation Counsel. It connects human contribution, patent inventorship, ownership, federal funding, and protection strategy without assuming that AI use created a problem.