A patent has a box marked ‘inventor’ that the law has always kept for a human. What happens the moment a machine’s name goes in it?
Stephen Thaler, an AI pioneer, believes his machine DABUS genuinely generated those gadgets on its own. Ryan Abbott, a law professor, had a sharper worry: if an AI designs a life-saving drug and the law says no human invented it, no one can patent it, and no company will pay to develop it. So they ran the Artificial Inventor Project.
In July 2021, Justice Beach became the first judge anywhere to rule that under the Patents Act a machine could be named as the inventor (South Africa had waved one through days earlier, but that was a formalities check, not a court weighing the question — so the world-first ruling was Australia’s).
It didn’t last. In April 2022, a full bench of five judges overturned him: an inventor must be a natural person. In November the High Court refused to hear it. And everywhere else the question was put — the UK, the US, Europe — the answer came back the same. No.
A machine can generate an invention, but under the law an inventor — like an author — has to be a person. Australia’s brief ‘yes’ is the world-first, but the settled answer everywhere is no.
The cases the traditional account teaches are all from before AI existed — Cantarella (2014, trade-marking the Italian word for gold on coffee), Seafolly (a designer’s Facebook post), Apple v Samsung (rounded corners). Meanwhile the fights that will actually decide this are unresolved and happening now: The New York Times v OpenAI, Getty v Stability.
And the same answer holds for authorship: an AI cannot hold a copyright either. The US said so in Thaler v Perlmutter; Australia had said it years before. An author, like an inventor, has to be human.
What was the DABUS / Artificial Inventor Project trying to force?
What did Australia decide, and does it stand?
Can an AI hold a copyright?