While the artificial intelligence (AI) application ChatGPT going mainstream has engendered thousands of column inches and terabytes of content in the physical and digital worlds respectively, in the back offices of academia and the courts, a new flurry of activity has emerged concerning the question of who—if anyone—owns the rights to outputs produced by these revolutionary new technologies.

Determining ownership is important if these new technologies are to be used for society’s maximum benefit. As has been seen with past advances in digital technology, if ownership is uncertain, then there will be too little “good” (productive and beneficial) creation and too much “bad” (illegal copying) creation will arise from their use. Also, the question of who will be liable for any negative consequences from AI creations needs to be resolved.

Take, for example, the website Character.AI. On this site, users can create characters and talk to, or “chat,” with avatars representing a variety of historical or fictional “pre-prepared” characters, ranging from the likes of Socrates and Queen Elizabeth II to Yoda, Superman, and Elon Musk. It is certainly amusing for a user to learn Yoda’s view that Luke Skywalker is the greatest Jedi of them all and that Her (late) Majesty’s message from beyond the grave is that she loves her grandson dearly but he should not have published that book. However, it is not quite so funny for Disney Entertainment as the Star Wars copyright owner or for Buckingham Palace as the custodian of the late Queen’s more refined memory.

It seems to be only a matter of time before a tsunami of cases lands in the courts as legitimate rights holders seek to determine the extent to which such creative uses by AI constitute infringements.  Earlier cases—such as Getty suing Stability AI for scraping its online images and a class action suit against Stability AI, DeviantArt, and Midjourney for training their models with billions of copyrighted images—have focused on the use of inputs to AI processes. However, new issues are arising from the outputs of AI engines. Has Character.AI infringed by allowing particular character avatars drawing on copyrighted material to be created using its software in the first place? Or is the individual who input the prompts to create the likes of AI Yoda at fault? What responsibility does the platform or the user bear for publicly sharing royal misinformation?

What is not in doubt is that the stakes are high and the legal cases are complex, as illustrated by an experiment by a University of Surrey law professor. Professor Ryan Abbott has been testing patent law in various jurisdictions around the world to see if an AI program’s inventive output could be protected in the absence of a human inventor. DABUS (Device for the Autonomous Bootstrapping of Unified Sentience), developed by US-based physicist Stephen Thaler, is the subject. In May 2021, Abbott applied to the New Zealand patent office for DABUS to be listed as the sole inventor of a food container it created and recognize that Thaler, as DABUS’s “owner,” was entitled to the invention. (That is, the company that owns an inventive AI is entitled to the patents for its creations. More generally, the company owning a creative AI is entitled to the intellectual property in the outputs it creates.) Similar applications were made in other jurisdictions.

The Intellectual Property Office of New Zealand declined to issue the patent, ruling that the term “inventor” refers to a natural person, which DABUS is not. An appeal to the High Court was dismissed. Lawyers for Thaler (and DABUS?) argued for the actual deviser to be named as the inventor regardless of whether or not that deviser was human: “The Assistant Commissioner’s interpretation limiting ‘inventor’ to a natural person and not extending the definition to a non-human entity is plainly erroneous.” If AI was not recognized as an inventor of what it invents, then those inventions would go unpatented, thereby discouraging the use of AI for innovation. However, the judge ruled that “it is fair to say that the natural reading of the section suggests the inventor is a person.” There was nothing to indicate that Parliament intended to own up the possibility of AI being an inventor when New Zealand’s Patents Act 2013 was passed, and it was not the role of the courts to do so.

So far, only South Africa has recognized DABUS as an inventor. The UK court came to a similar finding as New Zealand, but its ruling—along with 11 other cases from the experiment, including the US—is either under appeal or still awaiting a decision.  

Clearly, as AI Yoda would say, “Straightforward, life is not, for creations of AI machines—humans the creators are.”

The post AI Created It, But Who Owns It? appeared first on American Enterprise Institute – AEI.