On 4 March 2026, Japan's Supreme Court dismissed the final appeal in the long-running DABUS patent case, confirming that under Japan's Patent Act, "an inventor must be a natural person" and that AI systems do not qualify. The ruling closes a seven-year global legal saga and, more importantly for enterprise IP strategy, it removes the last major uncertainty about whether any of the world's largest patent systems will recognise an AI system as an inventor.
They won't. But the ruling also sharpens, rather than resolves, the much harder question every AI-using R&D team actually faces: how much human involvement is enough when AI materially contributes to an invention?
The DABUS Case
Researcher Stephen Thaler began filing patent applications in 2019 naming an AI system he built, DABUS (Device for the Autonomous Bootstrapping of Unified Sentience), as the sole inventor of several inventions, including a food container design. Thaler's position was that forcing a human name into the inventor field, when the AI genuinely conceived the invention without human direction, amounted to a legal fiction.
Japan's Patent Office rejected the applications, triggering litigation through multiple court levels that concluded with this year's Supreme Court dismissal.
A Rare Global Consensus
What makes this moment notable isn't just the Japanese ruling, it's that Japan now joins a consistent bloc. The United States, United Kingdom, European Union, Germany, Australia and Canada have all rejected AI inventorship in DABUS-related litigation. Only South Africa's patent office accepted a DABUS filing, and legal commentators widely characterise that as a procedural quirk (South Africa's patent office does not conduct substantive examination) rather than a considered legal endorsement.
Eight major jurisdictions arriving at the same answer, through different legal systems and different reasoning, is about as close to global consensus as patent law gets.
Why: The "Conception" Standard
The rulings converge on a shared legal concept: conception, the formation, in a human mind, of a definite and permanent idea of the complete and operative invention. AI systems generate outputs, but patent law requires an inventor capable of forming and testifying to a complete idea "in the mind." That structural requirement excludes AI from inventorship regardless of how capable the system is.
What Remains Unresolved
None of these rulings address the much more common real-world scenario: a human researcher directs an AI system to explore a solution space, the AI generates candidate outputs, and the human selects and refines the most promising one. Courts have not yet drawn a clear line on how much direction, selection, or refinement is enough to establish human conception when AI does a substantial share of the generative work.
Practical Guidance for AI-Assisted R&D Teams
1. Document Human Conception Contemporaneously
Given the unresolved middle ground, the safest practice is rigorous, contemporaneous documentation: lab notebooks, design records, and prompt/output logs that demonstrate a human researcher directed the inquiry, evaluated the AI's output, and contributed the inventive judgment that turned a candidate output into a complete, operative invention.
2. Name Human Inventors, Even in Heavily AI-Assisted Work
Following the consensus, patent applications should always name the human researchers involved, not the AI tool. Where multiple researchers contributed differing amounts of conceptual input, standard co-inventorship rules apply exactly as they would in any human collaboration.
3. Set an Internal AI-Assisted Invention Policy
R&D organisations that use generative AI extensively in ideation should adopt a written internal policy defining what level of human review and modification is required before an AI-assisted output can be treated as a patentable, human-conceived invention.
Conclusion
The DABUS saga is over, and the answer is unambiguous: AI cannot be named an inventor, anywhere that matters commercially. The next seven years of patent litigation will not be about whether AI can invent. They will be about how much AI assistance an invention can absorb before a human inventor's claim to conception starts to look thin, a question this ruling deliberately leaves open.