If Androids dream of electric sheep, who owns the sheep?


The technological landscape has changed vastly over recent years, not least in the world of Artificial Intelligence: AI is now part of our daily life whether we are aware of it or not, from the algorithms behind news feeds on Facebook, to the friendly chat bots helping us decide what clothes to buy. More than that, AI is evolving to fulfil more creative roles: The Washington Post uses a “robot reporter” which has published over 850 articles and in 2016 the IBM Watson supercomputer created a trailer for the film Morgan. But as the role of AI grows, businesses need to become increasingly aware of the impact that the use of AI might have on their intellectual property.

The November Budget set out the intention to establish the UK as a world leader in innovative technologies such as AI, with the promise of more investment and a focus on developing research in the sector. However, the government report into the AI industry, released in October, made no suggestion as to how, or indeed whether, AI is to be regulated nor how its outputs will interact with intellectual property law.

Creative AI

AI refers to systems that can learn and adapt, extracting data using reasoning and logic to produce an output. Ultimately this will involve a system which uses autonomy and creativity, but we are still a long way off from a “true” AI capable of sentience. Nonetheless, creative works are beginning to be produced using AIs; recent examples include DeepMind’s generation of short piano compositions and “The Next Rembrandt”, an original artwork created using analysis of Rembrandt’s works to recreate technical and aesthetic elements, producing something which could have been created by the Master himself.

The purpose of intellectual property law is to protect the product of intellectual endeavour and creativity. If AI is capable of producing creative works then the question arises as to who owns the intellectual property in that work, and whether it is even possible or desirable to protect intellectual property created purely by an AI.

Copyright protection gives the owner the exclusive right to copy, perform, show or adapt the protected work. Without this right third parties can make free use of the work and its value is diminished. The Copyright, Designs and Patents Act 1988 (CDPA) states that ownership of any copyright rests with the author, or the author’s employer, where the author is “the person who creates” the relevant work. AIs do not have legal personality and will not be considered employees, therefore this raises concern that if there is no human author then any output might be free from copyright. The same issue arises with design rights and patents: in order for the right to exist there must be a human author or inventor.

The recent US case of Naruto v Slater, concerning the “monkey selfie”, considered whether a non-human could hold copyright. The court found no copyright resided in the picture because a human had not taken the photograph. Although no similar cases have been heard in the UK, this may suggest that if there is truly no human input courts may be inclined to rule that there is no copyright. Uncertainty in the ownership and related rights of AI-created works may result in unwillingness from businesses to invest large amounts of resources into AI technology, reducing progress in the field.

One possible solution to this challenge may be found in the CDPA’s provision for the copyright of computer-generated works in which there is no human author. In such instances the author is the individual who makes the necessary arrangements for the creation of the work. However it is far from clear as to what constitutes “necessary arrangements” and whether this would include the creator of the AI or the person who puts it to use. Further, this provision was drafted long before AI, at a time when there was a fairly direct link between the human input and the computer-generated output. Modern AI systems are several orders of magnitude more complicated, and can draw on huge swathes of data, which can produce results that are both more unexpected and apparently “creative” than the humans who created the AI imagined.

Whilst the above has not been explicitly dealt with, case law may provide some assistance. Nova Productions v Mazooma Games considered the impact of player input into a video game in the context of ownership of the resulting copyright. It was found that the player’s input was not artistic, nor did the player contribute any artistic skill or labour, therefore the copyright belonged to the designer of the game who had undertaken the necessary arrangements for the creation of the work. Therefore, if a third party makes use of the AI system they would need to make an artistically skilled contribution to the work in order for them to have any claim over the copyright.

However it is not clear that generally the creator of an AI would have undertaken the necessary arrangements to gain authorship of the output. In order for a work to enjoy copyright protection a minimum amount of effort must have been expended in its creation. Whilst the bar for effort is low, it is nonetheless a bar. In Ray v Classic FM it was found that in order for a joint author to gain ownership that author must have actively participated in the creation of the work, and contribution of ideas alone was not sufficient. Here a distinction can be drawn between the creation of the AI and the creation of the work itself. This is possibly what copyright ownership may come down to - whether the AI is merely a tool used by a human to create something, or whether the creative spark required for intellectual property comes from the AI itself. In the case of the latter, the law as it stands provides no protection for the works created.


The current position of intellectual ownership and AI is uncertain, and there is the risk that if adequate protection for AI-created works is not provided then investment and innovation will decrease, as AI become more capable of creating works without human input. AI may not yet be able to independently produce work requiring protection, but with the current rate of technological progress this day may not be far off; the law will need to adapt as the need for human input and creativity diminishes.

Tom Lingard, Partner and Head of Intellectual Property, Stevens & Bolton
Camille Arnold is a Trainee at Stevens & Bolton
Image Credit: CyberHades / Flickr