The UK government and US District Courts have rejected the suggestion that AI systems can own copyright. Schellion Horn and Tom Middleton explain the valuation implications of intellectual property rights for AI-created outputs.

AI can support innovation and creativity. For example, it can be used as a tool by scientists, entrepreneurs and artists, and is increasingly being embedded into every element of our lives. Some believe that AI has begun to invent and create in a way that makes it impossible to identify specific human intellectual input in the creation. If this is the case, does an AI system have intellectual property rights over the outputs it creates?

The UK government framework

In 2021, the UK government sought evidence and views on a range of options on how AI should be dealt with in the existing patent and copyright systems. The consultation considered three potential outcomes:

1 Copyright protection for computer-generated works (CGWs) without a human author

2 Licensing or exceptions to copyright for text and data mining (TDM), which is often significant in AI use and development

3 Patent protection for AI-devised inventions

The government concluded that in the case of CGWs any change in the law could have unintended consequences since the use of AI is at early stages and a proper evaluation of options is not possible. For AI-devised inventions, the government noted that most respondents felt that AI is not yet advanced enough to invent without human intervention and as such no change to UK patent law to recognize AI ownership was needed. Although the UK government noted that it would keep this area of law under review to ensure that the UK patent system supports AI innovation and that it will seek to advance AI inventorship discussions internationally to support UK economic interests.

Testing of the US copyright rules

Since UK government's consultation, Stephen Thaler (a US citizen) designed an AI system – known as the 'creativity machine'. This 'machine' produced a work of visual art which Thaler argued was created autonomously by the AI system without human involvement. Thaler tried to have the AI system registered as the owner of the copyright in the US. The US District Court for the District of Columbia (in Thaler vs The register of copyrights) rejected the AI system as being the copyright owner and reaffirmed its position that human authorship is a pre-requisite for copyright protection.

Had Thaler not been testing the legal approach, he could perhaps have argued that he had copyright since he developed the AI system that had created the work of art – that is, the AI-system was supportive of his creativity, but not its sole inventor - hence the human authorship was his through the system development process.

In the UK, Thaler has been seeking patent protection for two inventions that he states were created by another AI 'machine' DABUS. The Comptroller-General of Patents, Designs and Trademarks refused to designate DABUS as the inventor as it is not a person. Thaler's initial appeal was dismissed by the High Court and the Court of Appeal. Thaler is appealing to the Supreme Court. The hearing for this appeal took place in March and is currently awaiting judgement. 

Learn more about how our Dispute economics services can help you
Economic and financial expertise in litigation and arbitration
Learn more about how our Dispute economics services can help you
Visit our Dispute economics page

The other side of copyright in artificial intelligence

AI systems learn from data. The data that an AI system learns from is referred to as training data. What was not dealt with in this trial was the copyright status of the training data on which AI systems rely. Thaler’s 'creativity machine' relies upon artificial neural networks which are trained on data. It is unclear the extent to which this training data may be subject to copyright protection.

Separate proceedings in a different US District Court against OpenAI and Meta brought by comedian Sarah Silverman and authors Christopher Golden and Richard Kadrey allege that OpenAI’s ChatGPT and Meta’s LLaMA were trained on illegally acquired datasets containing their copyrighted works.

Implications for the value of AI-generated outputs

While the decision in Thaler’s case is expected to be appealed, this decision itself has important implications. AI is increasingly used in creative processes and the value of both tangible and intangible AI created assets will depend upon intellectual property rights, including patents and copyright. They must provide the right incentives to AI development and innovation. They must enable the innovation to be appropriately owned and valued such that they can be given an economic value and traded.

Certainty regarding the economic incentives arising from the use of generative AI across the economy, but particularly in more artistic and creative applications, is important in the promotion of investment and adoption. Such certainty will only prevail when the copyright status of training data and generated content is resolved.

Learn more about how our Regulatory economics and finance services can help you
Economic analysis and advice for regulated sectors
Learn more about how our Regulatory economics and finance services can help you
Visit our Regulatory economics and finance page

For more insights and guidance on valuing AI systems and their outputs, get in touch with Schellion Horn or Tom Middleton.