The question of whether artificial intelligence can be recognized as an inventor under patent law has moved from theoretical debate to active litigation and legislative consideration. As AI systems become increasingly capable of generating novel solutions, materials, and designs with minimal human guidance, the legal frameworks built around human inventorship are being tested in unprecedented ways. The outcome of this debate will shape innovation incentives, patent strategy, and the broader relationship between law and technology for decades to come.
The DABUS Cases: A Global Test
The debate crystallized around DABUS (Device for the Autonomous Bootstrapping of Unified Sentience), an AI system created by Stephen Thaler. Thaler filed patent applications in multiple jurisdictions listing DABUS as the sole inventor for two innovations: a fractal-geometry food container and a neural-flame light device. The applications were designed as test cases to force patent offices and courts to address AI inventorship directly.
The results varied dramatically across jurisdictions. The United States Patent and Trademark Office (USPTO) rejected the applications, and the Federal Circuit upheld the rejection in Thaler v. Vidal, holding that the Patent Act's use of the term "individual" unambiguously refers to a natural person. The court emphasized that patent law is a statutory creation and that any expansion of inventorship to include AI systems must come from Congress, not judicial interpretation.
European Patent Office
The European Patent Office (EPO) similarly rejected the DABUS applications, finding that under the European Patent Convention, an inventor must be a natural person. The EPO's Board of Appeal affirmed that designating a machine as an inventor does not meet the requirements of the Convention, regardless of the machine's capabilities. However, the decision left open the possibility that patent applications for AI-generated inventions could succeed if a natural person is named as the inventor.
Divergent Outcomes
Not all jurisdictions agreed. South Africa granted a patent listing DABUS as the inventor, though South Africa's patent system does not conduct substantive examination, limiting the precedential value of this decision. Australia's Federal Court initially ruled that AI could be an inventor under Australian law, but the Full Federal Court reversed this decision on appeal, aligning with the US and European positions.
Why the Inventor Question Matters
The debate over AI inventorship is not merely academic. It has practical implications across several dimensions of patent law and innovation policy.
- Ownership and assignment: Patent rights flow from the inventor. If an AI cannot be an inventor, questions arise about who owns rights to AI-generated inventions and whether they can be patented at all.
- Disclosure incentives: The patent system exists to incentivize disclosure of innovations in exchange for time-limited monopoly rights. If AI-generated inventions cannot be patented, innovators may resort to trade secret protection, reducing public access to knowledge.
- Validity challenges: If a human is named as the inventor of an AI-generated invention but did not actually conceive it, the patent could be vulnerable to invalidity challenges based on misrepresentation of inventorship.
The Human Contribution Framework
Most patent offices have converged on a practical approach: AI-generated inventions can be patented as long as a natural person who made a meaningful intellectual contribution to the invention is named as the inventor. The critical question becomes what constitutes a sufficient human contribution.
The USPTO issued guidance in February 2024 clarifying that while AI cannot be named as an inventor, the use of AI in the inventive process does not automatically disqualify a patent application. A natural person can be listed as an inventor if they made a "significant contribution" to the conception of the invention. The guidance identifies several factors relevant to this determination.
Factors for Significant Contribution
According to the USPTO framework, a person may qualify as an inventor of an AI-assisted invention if they constructed the prompt in a manner that elicited a particular solution, recognized and appreciated a novel output from the AI, or took the AI's output and applied their own expertise to reduce it to practice in a way that required inventive effort. Merely owning or operating an AI system, without more, is insufficient.
This framework creates a spectrum. At one end, a researcher who uses AI as a sophisticated tool while maintaining close intellectual control over the inventive process clearly qualifies as an inventor. At the other end, someone who inputs a generic prompt and receives a fully formed invention without understanding or modifying it likely does not. The gray area between these extremes will generate litigation for years to come.
Patent Prosecution Strategy for AI-Assisted Inventions
Given the current legal landscape, organizations using AI in their R&D processes should adopt deliberate patent prosecution strategies. Documentation of the human contribution throughout the inventive process is essential. This includes recording the problem formulation, the design of prompts or parameters provided to the AI system, the evaluation and selection of AI outputs, and any modifications or refinements made by human researchers.
Patent applications should clearly articulate the human inventor's contribution in the specification and prosecution history. Vague or formulaic descriptions of the inventive process may invite examiner scrutiny or future validity challenges. Some practitioners recommend including a brief description of the AI tool's role and the human inventor's specific intellectual contributions in the application itself.
Trade Secret Considerations
Organizations that determine their AI-generated innovations do not meet the threshold for patent protection may consider trade secret protection as an alternative. Trade secret law does not require identification of a human inventor, and the protection lasts as long as the information remains secret and provides competitive advantage. However, trade secrets are vulnerable to independent discovery and reverse engineering, and they do not provide the offensive enforcement capabilities that patents offer.
International Harmonization Efforts
The World Intellectual Property Organization (WIPO) has convened multiple sessions to discuss AI and intellectual property policy, including the inventorship question. While no consensus has emerged, several proposals have been floated, including creating a new category of protection for AI-generated inventions, requiring disclosure of AI involvement in the inventive process, and establishing international guidelines for assessing human contribution to AI-assisted inventions.
The lack of international harmonization creates strategic considerations for multinational patent portfolios. An invention that qualifies for patent protection in one jurisdiction may not in another, depending on how each jurisdiction assesses the sufficiency of human contribution and defines inventorship.
Future Directions
Legislative reform remains the most likely path to resolving the AI inventorship question definitively. Several countries have initiated consultations or reviews of their patent laws in light of AI developments. Any reform will need to balance multiple policy objectives: encouraging innovation, ensuring adequate disclosure, maintaining the integrity of the patent system, and addressing the practical realities of increasingly autonomous AI systems.
Until legislative clarity emerges, organizations should focus on documenting human contributions to AI-assisted inventions, developing internal policies for AI use in R&D, and monitoring legal developments across their key patent jurisdictions. The companies that adapt their patent strategies proactively will be best positioned to protect their innovations regardless of how the law ultimately evolves.
