The AI Act and Intellectual Property Rights: Key Obligations for Providers of General-Purpose AI Models
The European Union’s Regulation 2024/1689, commonly referred to as the AI Act, represents a pioneering approach to regulating the deployment of artificial intelligence tools. This legislation categorizes AI systems based on their risk levels and imposes corresponding obligations on their providers. Below, we explore the key obligations that the AI Act places on providers of general-purpose AI (GPAI) models, particularly in relation to intellectual property rights.
1. Copyright Policies Regarding General-Purpose AI Models
Providers of general-purpose AI models are required to implement copyright policies that demonstrate compliance with EU laws on copyright and related rights (Art. 53.1(a)). A general-purpose AI model is defined as an AI system capable of performing a wide range of tasks, trained on large datasets through self-supervision.
Notably, AI models that are used solely for research, development, or prototyping prior to market release are excluded from this definition. The copyright policy must illustrate how providers will respect the rights reserved by copyright holders. Under the European Copyright in the Digital Single Market Directive (Directive 2019/790), an exception for text and data mining allows GPAI providers to reproduce or extract protected content for analytical purposes, provided that rights holders have not reserved their rights in a clear manner.
Importantly, the AI Act confirms that the text and data mining exception applies to GPAI providers, who must now document their compliance with copyright reservations effectively.
2. Information Policies for AI Systems Implementing General-Purpose AI Models
According to Art. 53.1(b) of the AI Act, GPAI providers must offer up-to-date information policies for customers integrating their models into AI systems. These policies must furnish users with a comprehensive understanding of the model’s capabilities and limitations. Essential elements to include are:
- Details of the tasks the model is designed to perform.
- An outline of acceptable use policies.
- Information on interactions with external hardware or software.
- The architecture of the model and its parameters.
- Version details of relevant software.
- The modality (text, image, etc.) and formats of inputs and outputs.
- The licensing terms for the model.
- The release date and distribution methods.
These policies should be aligned with the provider’s own intellectual property rights and confidential business information.
3. Training Data Policies Regarding General-Purpose AI Models
GPAI providers must also outline a training data policy that details the data used for (pre-)training their models (Art. 53.1(d)). This policy allows rights holders to ascertain whether their data was utilized in the training process. Key components should include:
- Data collections employed during training.
- Permissions obtained from data providers.
- Measures to protect trade secrets or confidential information.
The EU body responsible for enforcing the AI Act is expected to ensure compliance with training data and copyright policies, potentially providing templates to streamline this process.
4. Conclusion
The AI Act represents a significant advancement in the regulation of artificial intelligence, aiming to ensure that AI development adheres to stringent standards. By mandating transparency related to copyright compliance, training data policies, and more, the EU intends to safeguard intellectual property rights while fostering innovation within the AI sector. As the AI landscape evolves, the provisions set forth in the AI Act will be instrumental in promoting responsible and ethical AI practices.
