It is always interesting, at least for me, to observe the emergence of new technology and its interaction with the current legislation and new approaches from both academic and practical perspectives. Here is the summary of recent studies and opinions on the topic of copyright regulations related to AI-based systems, especially thanks to the blog article by M. Matt Hervey.
Image by Gerd Altmann from Pixabay
AI-generated works
This is not an easy question to find the most persuasive, efficient answers. The discussion focuses on the copyright law principles, such as authorship – human-centred authorship requirements, protection conditions – notions of creativity, originality, dialectics between privatisation and public sector, the rationales, subsidiarity and proportionality of copyright legislation.
There are a few countries acknowledging the copyright on computer-generated works. Their copyright regulations define the author who is the person undertaking arrangements necessary for creating the work. Hong Kong Section 11(3) Copyright Ordinance , Ireland Art. 21 Copyright and Related Rights Act 2000, New Zealand Section 5(2)(a) Copyright Act 1994, South Africa Section 2(h) Copyright Act 1978, the UK Section 9(3) Copyright, Designs and Patents Act have the same wording. Or, in other words, the author is the person causing the work to be created – India Chapter I Section 2(d)(vi) Copyright Act, 1957.
In 2019, the U.S. Copyright Office (USCO) refused the registration of AI-generated artworks, including DABUS’ image A Recent Entrance to Paradise. Then, it recently doubted its own approval decision for the registration of MidJourney’s graphic novel Zarya of the Dawn. The human cause is the main issue of these discussions.
Text and data mining
The copyright law provides a very thin protection layer for databases in addition to the sui generis right. However, if the text mining reproduces and extracts the data which is copyrighted work without the author’s consent, it shall cause an infringement unless the regulated exceptions otherwise permit it. For example, many countries have adopted these exceptions, for example:
- The U.S. Copyright Act 1976 Section 107 and several cases law (Authors Guild v. HathiTrust, White v. West, etc.) uphold the text mining as fair use.
- The UK Section 29 A Copyright, Designs and Patents Act 1988 allows non-commercial text and data analysis exceptions.
- The EU Directive on Copyright in the Digital Market supports text mining within certain limitations.
- Following the EU Directive, France L.122-5-3 Code de la Propriété Intellectuelle grants this exception for the scientific research purpose.
- Japan Article 30-4 Copyright Act also favours this exception provided that the process does not involve what the work expresses and human senses perceive that work.
- Singapore Copyright Act 2021 permits the exception of copying of copyright works for purpose of computational data analysis, including text mining, training machine learning and sentiment analysis.
Caricature, parody or pastiche
The InfoSoc Directive Article 5.3(k) in 2001 introduced this exception which is relevant to the freedom of expression. Up to now, there has not yet been a case law so that we can observe the consequent impact of the application of this exception. On the one hand, it is a very promising way-out for AI-generated works, imitating those created by humans. On the other hand, this will pose the challenge to examine and evaluate the criteria to apply this exception.
Other scenarios…
As many governments’ concerns, the AI technology application is at its early blooming stage, we have not gathered enough resultant effects to society and economy. The EU Intellectual Property Office (EUIPO) published last year its study on the impact of AI on the infringement and enforcement of copyright and designs of physical products, digital content. Through the fictional scenarios, they estimated the potential infringements which appear to be real risks for the right-holders.
For the digital content, they envisaged the following storylines: hacking media accounts, computer icon/ virtual commodity infringement, social media offences, sales of hacked media accounts on the dark web, media sharing platform offences, virtual/ gaming world offences, P2P and BitTorrent, live streaming – sports broadcast link aggregator (ads-based), IPTV crime – unauthorised access to subscription-based IPTV service, AI training. In terms of physical products, the scenarios are more about physical/ online marketing of copyright & design-infringing products. In addition, the EUIPO also considers the theft of copyrightable work, design under development, design registration fraud, mass production and importation of copyright and design-infringing goods.
Conclusion
In sum, the AI systems usually exploit an enormous amount of data for designing, training, and deploying; therefore, it is not simple to extract the data and prove its origin and copyright infringement. We must wait for both technology and legislation to have a satisfactory conclusion.
N.B. I do wish to write more articles with higher quality and eager to learn more from the readers. Thus, one, I’ll open the comment section with hope to obtain your feedback and discussion. Please be nice and do not attack my little site. Second, I’ll rearrange the posting schedule to invest more in the analysis and writing.
For more topics about copyright covering AI-based systems on AstraIA Gear: ChatGPT: what legal and ethical issues do you need to know?, Weekly Legal Updates in Tech – U.S. lawsuits, How to use AI to protect your intellectual property rights
For more short news follow me, or our Page on LinkedIn, cheers!