Intervention to WIPO’s AI and IP conversation
The WIPO started an important consultation on the intersection and a possible regulation of AI and IP. Following an open consultation phase, the WIPO is planning to have an online conversation in July. As a part of that, interested people might be able to submit interventions and they might be allowed to speak on the event. I also submitted my intervention.
This document is purposefully short – my opinion is generally public, included in a draft paper accepted for publication by UFITA. The full text of my intervention is as follows:
June 23, 2020
following the call for interventions to the “WIPO Conversation on Intellectual Property and Artificial Intelligence”, and reflecting on Issue 7 of the Revised Issues Paper on Intellectual Property Policy and Artificial Intelligence, I hereby intervene to argue against the protection of AI-generated outputs under the existing copyright norms (including neighbouring rights and sui generis regimes).
Issue 7 includes Question 23, which lists eight sub-questions. The present intervention argues that sub-question (i)-(vi) and (viii) should be answered to the negative in the sense that copyright (including neighbouring and sui generis) protection shall not extend to outputs generated by AI, where the causal link between the original human input and the AI-generated output breaks, that is, there is no direct human control over the creation of the final output.
This opinion can be supported by doctrinal/historical reasons, policy considerations and the lack of empirical evidences to support the need of copyright protection.
(1) Doctrinal/historical reasons include (but are not limited to)
- the historical development of copyright (copyright law has always rewarded humans for their creative activities);
- the incentives connected to copyright law (all relevant incentive theories, that is, the personality, the labour or the utilitarian theory, are inherently based the human “behind” the work);
- the concept of authorship (even in the lack of an express definition of author in international – and most often on domestic – level, copyright norms include solely humans in the concept of author);
- the concept of originality (which is bound to the concept of author and subject matter, as well as in various jurisdictions to creativity, and it also reflects the need for “intellectual” activity – rather than any shady concept of “intelligence”);
- and moral rights (while the scope of moral rights differ in the various jurisdictions, their essence remains the same everywhere: it reflects the personal connection between the human author and her work).
Domestic copyright norms regulate some of the above elements differently, and, in various cases, they also include notable exceptions to the above (especially related to copyright ownership, e.g. under the work-made-for-hire doctrine). Still, no existing copyright regimes could cover AI-generated outputs without doctrinal inconsistencies or imbalances between human-created works and AI-generated outputs.
(2) Other than the incentives behind copyright law, as introduced above, further policy arguments speak against copyright protection of AI-generated outputs. Copyright law – in its purest sense, focusing on authors – is not an investment protection scheme. While neighbouring rights and sui generis protection focus on investment interests, too, the original policy purposes related to them differ significantly from those that might be relevant for AI creators or investors. There is no convincing policy legitimation of the protection of AI-generated outputs besides or against the beneficiaries of the regular copyright system. Indeed, the protection of AI-generated outputs besides or against the human rights holders by the same (already existing) copyright regime simply does not worth if the AI-industry could harm the individual human interests, the cultural interests of the society and the creative industry in general.
(3) There is a visible and serious lack of empirical evidences to support the need of copyright protection for AI-generated outputs. The lack of evidences are related to
- the (global, regional or domestic) volume and value of AI-industry and its effects on the promotion of culture and economy;
- the positive or negative externalities of any protection (otherwise: could the lack of protection harm the AI-industry and, vice versa, can the protection of AI-industry harm the “regular” forms of copyright industry?);
- “AI winters”: will any AI winter surface at all and would any AI winter surface due to the lack of copyright protection;
- the market/need for AI-generated outputs in all or any segments of cultural industry.
In sum, the present intervention argues that AI-generated outputs simply do not fit into the concept of existing copyright norms, and those norms cannot and should not be flexibly interpreted to cover such outputs.
Based on the above, and in response to sub-question (vii) of Question 23, the present intervention argues that the maximum protection that AI-generated outputs might receive is a new sui-generis regime,
- which shall neither replace nor be equal with the copyright regimes protecting human creativity and/or contribution;
- the scope of which can only be set following a deep empirical and multi-disciplinary analysis of the marketability of outputs (economic analysis) as well as the individual and cultural/social consequences (cultural analysis) of an AI-centred protection upon an human-centred regime, where the ultimate goal shall remain that copyright law aims the development of individual humans and the society/humankind in general.
This intervention is drafted to reflect and comply with the strict conditions of a possible oral intervention. My arguments on this issue might be available in its full version via https://ssrn.com/abstract=3592187.
Dr. habil. Péter Mezei
Associate Professor, Institute of Comparative Law and Legal Theory, Faculty of Law and Political Sciences, University of Szeged
Adjunct Professor (dosentti), University of Turku