Artificial Intelligence
With the advent of generative artificial intelligence (GAI) systems, we introduced our AI principles last year to guide discussions and policymaking. These principles set out key considerations and emphasized the importance of a comprehensive legal framework that addresses copyright, privacy, data protection, and competition and consumer protection laws. Such a framework is essential for the responsible regulation of AI technologies.
As GAI continues to rapidly evolve, we believe it is increasingly important to address the potential consequences of GAI on the invaluable contributions that authors make to society, especially at the international level.
Context
The market for GAI systems, particularly those trained on existing content, is becoming increasingly commercialized. OpenAI’s transition to a for-profit model, backed by substantial investments from Microsoft, has propelled its generative AI systems like GPT-4 and DALLE into the creative industry, leading to the OpenAI’s valuation reaching $118 billion by October 2024. These systems rely heavily on written works, placing authors at the heart of the ongoing discussion about the use of their content in training GAI systems. As these technologies continue to evolve, it is essential to address the implications for authors whose works are being used without their consent or knowledge.
Clarity and certainty
It is important to distinguish between the use of copyrighted works for training GAI systems and other practices, such as Text and Data Mining (TDM), which were established before the rise of GAI technologies. To ensure clarity and certainty in regulation, a specific policy approach is needed to address the use of copyrighted works in AI training. The EU’s current opt-out system for copyright holders serves is an example that presents significant challenges. This system requires authors to actively opt-out of having their works used, which imposes an unreasonable burden on them in monitoring how their works are being used in AI training. It places the responsibility of enforcement on authors, who typically lack the resources or capacity to track unauthorized uses of their works, making the system both inefficient and impractical.
Transparency and Choice
The current GAI view has been imposed on authors whose works have been used by commercial entities without their consent, often without their knowledge of the extent of the use. It is therefore vital to establish a workable, regulated system that identifies the specific works of authors used within GAI systems. Greater transparency in author contracts regarding the past and future use of GAI rights will help create a more efficient and fair market based on choice rather than constraint.
For it to be effective, transparency must be to a high standard. Proposals from the tech industry have often opposed this, but authors need to be able to understand in detail where their work has been used.
Licensing options must ensure authors continue to have choice over the use of their works. Approaches such as an opt-out system would give authors only a binary choice of having their work used or not but, as with other areas of copyright, it is important that authors retain some control over the licensing of their works with a view to ensuring that the end use does not undermine their interests.
AI training rights are a recent and novel concept. It should not be assumed that authors previously signed over these rights in contracts agreed when such uses were not conceived. Greater transparency in authors’ contracts regarding past and future uses ofGAI rights is needed.
Looking at the future
Market-based solutions offer the most effective path for advancing GAI development without neglecting the rights and contributions of authors and creative industries globally. This market is developing through a variety of licensing models, and its foundation must be built on transparency regarding which works are being used and which are available for use. Such transparency will enable the development of legitimate GAI technologies and ensure fair licensing practices.
Conclusions
In conclusion, we believe that establishing clear international legal frameworks to regulate the use of copyrighted works in the training of GAI systems is essential. These frameworks should promote transparency and fairness, ensuring that authors’ rights are respected on a global scale. By supporting the development of a licensing market based on certainty, clarity, and choice, such regulations will foster the responsible growth of GAI technologies worldwide, while safeguarding the right of authors and the interests of the creative industry at international level.
