Creators and publishers are amongst those most affected by the development of generative AI models. Their works are being used to train AI systems without their authorisation, remuneration, or recognition, frequently under the guise of ‘research.’ Creators and publishers are also concerned about AI-generated works being passed off as works of authentic human creativity—a concern shared throughout society.
We acknowledge that there are useful and important purposes to which AI can be applied. However, there is an urgent need for policymakers to implement regulatory regimes to respond to these concerns. Creators and publishers should, we believe, participate in policy discussions to ensure that AI systems are transparent, ethical, fair, and lawful.
The decision to regulate or not and, if so how, is not a zero-sum choice. AI can fuel innovation, economic growth and productivity and be of special value to science. Achieving these benefits must not be at the expense of the creative, aspirational and commercial rights of creators and publishers.
Creative and publishing rights: protection against AI exploitation
The scraping, analysis, and exploitation of vast quantities of data frequently occur without rightsholders’ lawful authorisation. These works have value, both economic and non-economic, and rightsholders should be able to authorise or prohibit use of their works, to receive acknowledgement for it, and to be compensated for it.
Māori intellectual property
The exploitation of Māori taonga, in its impact on Māori creators and traditional systems on intellectual property management, and the effect of AI outputs on Māori tikanga are also an issue. Te Tiriti o Waitangi and Wai 262 oblige the Crown to protect taonga and tikanga. Others can speak to these issues with greater knowledge than us, but we acknowledge their significance.
Collective management of licencing
The existing individual and collective licencing systems offer an efficient and well-established solution to the licencing of a wide variety of works, including for books, articles, other written materials, and visual art. These systems can readily accommodate the equitable use of rightsholders’ creative works by AI system and some responsible AI developers already use collective licences.
Text and data mining exceptions: guaranteed opt-out
New Zealand is a signatory to the Berne Convention, Article 9(2) of which is unequivocal that exceptions to copyright, such as TDM (Text and Data Mining) exceptions that allow AI systems to exploit creators’ works without authorisation or remuneration should be avoided. TDM exceptions should be narrow, remunerative, and respect the rights of creators.
Evidence-based policy
Copyright law has successfully adapted to many new technologies and information delivery systems and other challenges. To continue this success, we should not rush into changes without deep consideration of the needs of creators and publishers. There is no evidence of market failure to suggest that changes to Copyright law are needed to incentivise AI-related innovation.
Policy changes and existing markets
Derogations of the rights of creators and publishers need to be subject to the exceptions and limitations defined in the Berne Convention. The application of exceptions to respond to AI should be no different: exceptions must consider the integrity of the overall copyright system and should not harm existing and developing markets.
Education
There is an acute risk of AI systems competing for economic and non-economic rewards with the underlying copyright works that enabled the AI system in the first place.
Sound copyright policy will be effective when policymakers, rightsholders, and all users of copyright material have robust understanding of the legal implications of using copyright works and the importance of copyright law to society. This needs to include appreciation that the use of copyright works as inputs to AI systems, without authorisation or remuneration to creators and publishers can, particularly in commercial contexts, result in infringement.
Transparency: the key to fair AI practice
Copyright policies that respond to the ethical implementation of AI would provide for the transparent recording and communication to creators and publishers of the copyright works used by AI systems, and for what purpose. This approach is critical to support the identification and licencing of such works. AI operators need to be legally required to maintain and allow access to this information if they are to be accountable for activities and outputs that infringe the rights of creators and publishers.
Co-regulation
The interplay between AI and the law is complex. We believe that a co-regulation approach would best guide the development of the AI ecosystem in New Zealand, with an agency having backstop regulatory powers to approve codes of practice created by AI operators and the licencing collective agencies, impose a code if not jointly developed, and impose financial penalties for breach.
The best of international standards can guide a New Zealand approach that balances innovation with ethical and legal safeguards. Assessing and enforcing standards of quality and responsibility on the licencing collection agencies, e-developers and users of AI technologies will ensure that the economic value of intellectual property and data privacy protections, and ethical use can be achieved.
