AI Can't Be an Inventor: Japan's Supreme Court Draws the Line at Being Human

AI now designs novel drug structures, discovers new materials, and optimizes circuits. Which raises a strange question: if an AI genuinely invents something, who owns the invention? Japan’s Supreme Court just answered, and it did not hedge. An inventor, the court ruled, can only be a human being.

This is more than a legal footnote. It’s a landmark test of how far AI can push into territory we’ve always considered human, and where the law decides to plant its flag. So let’s walk through the ruling, and through the surprisingly messy way the rest of the world is splitting on the same question.

DABUS, the Deliberate Provocation

At the center of this fight is an AI system called DABUS, built by American researcher Stephen Thaler. The acronym stands for “Device for the Autonomous Bootstrapping of Unified Sentience,” which tells you everything about the ambitions baked into it.

Thaler claimed DABUS had, on its own, invented two things: a specially shaped beverage container, and an emergency light that flickers in a pattern mimicking neural activity to grab attention. Then he filed for patents on both across dozens of countries, and in the box marked “inventor” he wrote not his own name, but DABUS.

Here’s the crux. Thaler’s position was that honesty demanded it. “I didn’t invent these, the AI did, so the AI is the inventor.” It was a principled provocation, engineered to force patent offices and courts into a corner they’d rather avoid. And it worked. Country after country was dragged to the same uncomfortable question.

Japan Draws a Clean Line

Japan didn’t flinch. The heart of the ruling: under Japanese patent law, an “inventor” means a natural person — a living human being. Full stop.

The reasoning is elegant. The patent system offers inventors a temporary monopoly as a reward, and in exchange they disclose their invention to society. But that whole reward-and-incentive structure was built around humans from the start. An AI can’t own a right. It has no legal personhood to transfer one. The moment you write a machine’s name in the inventor field, the entire premise of the system collapses.

The court added one more note worth underlining. Whether AI should ever count as an inventor is not for judges to decide through case law — it’s a question for the legislature. Under existing statutes, only humans qualify. If society wants to change that, it should write a new law. It’s a deliberately restrained posture: the judiciary declining to legislate from the bench.

The World Has Split in Two

The interesting part is that Japan isn’t alone — and the world hasn’t landed in the same place. Because the DABUS applications went out on nearly identical paperwork everywhere, they’ve become a natural experiment. And the results diverge sharply.

Start with the countries that slammed the door and insisted only humans can invent:

• United States: The Federal Circuit ruled that under patent law an inventor must be an individual, meaning a human. The Supreme Court declined to hear the appeal.
• United Kingdom: The Supreme Court held unanimously that an inventor must be a natural person, and DABUS doesn’t qualify.
• European Patent Office and Germany: Both rejected the applications, requiring an inventor with legal capacity — a human.

Now the countries that cracked the door open:

• South Africa: The first jurisdiction on Earth to register a patent naming DABUS as inventor. But South Africa runs a formalities-only system that doesn’t substantively examine applications, so whether this counts as genuine recognition is very much up for debate.
• Australia: A first-instance Federal Court ruling that an AI could be an inventor made global headlines — and was then overturned on appeal.

Net result: the human-inventor principle is winning, and winning broadly. But the cracks are real, and those cracks are what keep this story from being over.

So What About Inventions AI Helped Create?

Here’s where people get it wrong. These rulings do not mean that AI-assisted inventions can’t be patented. They mean you can’t put the AI’s name in the inventor box.

Right now, countless researchers use AI as a tool to find new drugs and design new materials. In those cases, the human researcher is listed as the inventor, and the patents issue normally. The AI is treated the way a microscope or a piece of simulation software is — a powerful tool, nothing more.

The trouble is that the line keeps blurring. If a human just points the AI in a rough direction and the AI supplies the actual idea, can we still call that human the inventor? The US Patent Office has already aimed straight at this problem, issuing guidance that a patent is fine even when AI was used, so long as a human made a significant contribution. In other words, the question is shifting from “is the AI an inventor?” to “how much did the human actually contribute?” That’s where the real fight is heading.

What the Law Is Really Protecting

On the surface, Japan’s ruling reduces to a single line: AI can’t be an inventor. But underneath sits a deeper question. The patent system was, from the beginning, a promise designed to reward human creativity. So what happens to that promise when machines start doing a large share of the creating?

For now, courts around the world have held the line at being human. And they’ve kicked the harder decision — whether to move that line — over to society and its legislatures, where they say it belongs. So here’s the thought to sit with. If the day comes when an AI truly invents something, can we keep holding out with the owner’s chair deliberately left empty?