A United States court decision has affirmed that works generated by artificial intelligence are not authored by humans, and so cannot receive copyright protection. In a Washington DC court decision Thaler v Perlmutter, Judge Howell has found for the defendant who is essentially the United States Copyright Office.
The case was about Dr Stephen Thaler’s computer system “Creativity Machine”, that he designed to generate visual art. He sought to be recognised by copyright law as the owner of art generated by that machine, but the Copyright Office and now the District Court has refused to do so, on the basis that the visual art lacked human authorship.
This decision is the latest of many. Mr Thaler has been on a global quest to find a court willing to protect the creations of his A.I. machine, as part of his crusade to achieve recognition by courts of the right of A.I. inventors to register their inventions. As detailed by The Economist Mr Thaler’s personal history is fascinating and sometimes heart-wrenching. He is quoted as seeing a refusal to recognise inventions of A.I. machines as “speciesism … discrimination against machines”.
Mr Thaler’s more famous machine is the Device for the Autonomous Bootstrapping of Unified Sentience, or DABUS. As the name suggests, it is an autonomously operating neural network, able to generate technical solutions without being provided a corresponding task.
For him, the stakes are high and not just because of machine personhood. The views of some are that more progressive legal views of patent or copyright authorship, that would allow A.I. to be the inventor or author of intellectual property, may make a jurisdiction more attractive for the best designers to work. They might incentivise the pursuit of patents generated by A.I., which already turbocharged in the U.S. from 3,267 in 2017 to 18,753 in 2021. However, courts have not bended to these policy considerations. Mr Thaler has lost almost every case he has filed in a court.
Thaler v Vidal was his most high-profile battle, a 2022 appeal to the U.S. Supreme Court of the decisions of the lower courts that refused to order the issuing of a patent for an invention he said his A.I. system created. The U.S. Supreme Court has been known to issue divided rulings, but did not hesitate to reject this claim. Like the U.S. Patent and Trademark Office, a federal judge in Virginia, and the Federal Circuit court, the Supreme Court ruled that U.S. patent law required an inventor to be a human being in order to receive a patent.
Another notable attempt by Mr Thaler was in the U.K. Court of Appeal. The Court refused to order the issuing of patents for a shape of food packaging and a flashing light for emergencies, but not because of the artificial intelligence. Rather, it was because of a failure to name a person as the inventor on the relevant forms. Lord Justice Briss memorably called the case “an object lesson in the risks of advocacy being distracted by glamour.” These cases can turn not on headline controversies but on banal bureaucratic forms.
Nevertheless, Lady Justice Laing gave this guidance to designers: “If patents are to be granted in respect of inventions by machines, the 1977 [Patents] Act will have to be amended.” This appeared to be an advice in no uncertain terms to legislatures: if you want A.I. to register ownership of intellectual property, you must make a new law. Thaler appealed to the U.K. Supreme Court and judgment will come shortly.
Mr Thaler also sought intellectual property law protection in Australia, in a time before A.I. was in our daily headlines. The Federal Court in 2021 initially found in his favour. However, this decision was overturned by the Full Federal Court the following year. The Full Federal Court sat with a five-judge panel, unusually large for such an appeal which possibly signals the gravity of the decision. In 2022, the High Court of Australia refused to hear the case which left the Full Federal Court’s decision as the authority.
Courts are generally refraining from upending copyright and patent law as we know it in the face of a heady stream of artificial intelligence development. The subtext underneath the rulings might be summarised as: It is not our job to interpret the word “person”, “inventor” or “individual” differently to how we always have because of this brave new world of science and innovation – over to you parliaments and legislatures.
A German patent office initially refused the registration of a patent invented by DABUS, however the Federal Patent Court permitted a human to be named as the inventor on the relevant application forms as a person “who prompted the artificial intelligence DABUS to create the invention”. The Court rejected the notion that the legislation permits non-natural persons to be designated as inventors.
The South African Companies and Intellectual Property Commission is one of the few offices to have accepted the patent, although this decision appears to have been based on the application meeting formal requirements rather than withstanding tests of whether it should be revoked.
These decisions will surprise few intellectual property lawyers. Patent and copyright legislation differ but a very common element is a clear and obvious requirement for a natural person (a human being in non-legal speak) to be the named inventor or author of something in order for that thing to be protected under intellectual property law. In Australia, to obtain a copyright in respect of a work, the author of that work must be a human being.
What may be more fertile grounds for legal debate is the question of whether humans aided or assisted by A.I. are eligible for intellectual property protection. The U.S. Federal Circuit left this question open, while the German decision demonstrated an awareness of this distinction. In Aristocrat Technologies Australia Pty Ltd v Commissioner of Patents  HCA 29, the Australian High Court split 3-3 on a parallel question, which was whether a computer implemented-inventions in that case was a patentable invention under section 18 of the Patents Act 1990 (Cth). Two blocs of the Court tabled near opposite rulings, which leaves the law in a state of ambiguity but indicates that there is likely a point at which a computer-generated invention might be patentable.
Alert readers may have seen the headline of this article refers to a lack of protection by “current intellectual property law”. Governments may decide to legislate to give A.I.-generated works or patents protection. In an amici curiae for the unsuccessful U.S. Supreme Court case, scholars pointed out that A.I. has been creating patentable inventions since 1983 and that failure to protect these inventions may reduce the activity of those instructing or developing these technologies. Over to the lobbyists.
Thaler v Commissioner of Patents  FCA 879.
Commissioner of Patents v Thaler  FCAFC 62.
Thaler v Commissioner of Patents  HCATrans 199.