On 12 December 2019, the World Intellectual Property Organisation (WIPO) launched a public consultation on intellectual property and AI. The full consultation document can be downloaded here. Interested parties are invited to submit their comments to [email protected] by February 14, 2020.
Among the many questions upon which WIPO is inviting responses, some of the most interesting relate to the legal status of AI-created inventions, copyright works and deep fakes.
Patents: The consultation invites responses to the questions: should inventions created by AI be patentable? who should be named as the inventor in such cases? and who should own them? These are important questions given that many existing national patent laws will only allow inventions to be patentable if they disclose a named inventor who is a human being. For example, the UK IPO recently updated its formalities manual as follows: “An AI Inventor is not acceptable as this does not identify “a person” which is required by law. The consequence for failing to supply this information is that the application is taken to be withdrawn under s. 13(2)“. This follows on from recent patent applications for inventions conceived by DABUS (an AI system created by Dr Steven Thaler – www.artificialinventor.com). As AIs are used more and more to solve inventive problems, and to create inventions, the issue of who should be named as the inventor, and who should own any resulting patents in them will increasingly become questions for policy makers and the courts to resolve. Current provisions are likely to prove unsatisfactory and it seems that UK patent law is presently ill-equipped to deal with an AI-driven future. It seems fairly obvious to us that, for now, the law should be changed such that an AI ought to be capable of being named as an inventor, and the owner of patent rights in such inventions ought to be the legal owner of the AI. This of course leads to bigger philosophical questions of whether an AI should ever be legally entitled to enjoy property rights of its own, but these seem to us to be questions for a few years hence.
Copyright: Similar questions arise in relation to copyright law and policy. For example, consider the following:
- a personalised highlights programme of a Japanese J-League football match, featuring a football fan’s favourite team and players;
- the pop song, “Mr Shadow”; (https://www.youtube.com/watch?v=lcGYEXJqun8)
- the painting, “Portrait of Edmond de Belamy”, which was sold at auction in 2018 for $432,000.
What do each of these have in common? Of course, each of the works is of a kind that is capable of attracting copyright, and each is demonstrably commercially valuable (although that is perhaps questionable in the case of “Mr Shadow”). However, the most interesting factor that these works have in common is that each is authored by an AI. With more creative output likely to be AI-authored in future, WIPO is seeking the views of interested parties in helping it frame copyright policy around the following questions and issues:
- Should works created by AI be capable of protection as copyright works? In UK law, this question has been settled (at least to some extent) for some time in s.9(3) of the CDPA 1988: “In the case of a literary, dramatic, musical or artistic work which is computer-generated, the author shall be taken to be the person by whom the arrangements necessary for the creation of the work are undertaken.” However, as AI becomes increasingly autonomous the question of exactly who, if anyone, makes the arrangements necessary for creation could become hard to answer. And this question is even more troublesome for some legal civil law traditions, given the history of droit d’auteur copyright systems being based on the concept of copyright as a reward for culturally valuable expressions of human creativity;
- Should such works be protected by a new sui generis right?
- Who should be the first owner of copyright in an AI-created work? The AI itself? (which would necessitate the AI having a legal personality); the inventor/owner of the AI system that created it? or the operator of such system?
- Questions relating to infringement: because AI-created works are generally authored following a process of machine learning, in which the AI digests and processes vast quantities of third party-owned copyright works in order to influence and inform the creation of its own works, a question arises as to whether or not the use of a copyright work as part of such a data-set should give rise to infringement actions. Here, policy makers will need to consider whether or not it should be (or should continue to be) a restricted act for an AI to process a third party-owned copyright work as part of a data-set for machine learning purposes. Such a restriction could necessarily have a deleterious effect on innovation in the creative arts by AI systems. The EU has already taken some steps in this regard in its recent Directive on Copyright in the Digital Single Market (2016/0280/EU) with its limited (and in this author’s view not entirely adequate) exception from copyright infringement relating to use of copyright works for the purposes of text and data mining (TDM) by research and cultural heritage institutions.
Deep fakes: An incongruous but rather interesting area upon which WIPO is consulting concerns questions relating to deep fakes (which involve the use of technology to falsely represent the image and voice of a person in a fake video, designed to make it look like the person was really part of the scene portrayed). WIPO is consulting as to whom the copyright in a deep fake should belong, and whether or not the persons falsely portrayed in such deep fakes should be equitably remunerated in return for their portrayal. In this author’s view these questions are probably adequately addressed already in copyright laws internationally and the pernicious aspects of deep fakes – and the potential distress and loss caused to persons portrayed in them – are probably better dealt with by looking for redress in other areas of the law, notably defamation, malicious falsehood, data protection and laws relating to privacy (and where available – such as in the USA – rights of publicity). To the extent that such laws are deficient in tackling the phenomenon of damaging deep fakes, then policy makers should look to strengthen these areas rather than relying on copyright as the answer to these problems (since it is likely that the person featured in the video will not be the owner of the copyright in the footage that has been assembled / created in the production of the deep fake).