News & Events

Trilligent hosted another Tech Talk event in Brussels on October 28, 2024. Our Brussels event explored intellectual property in the age of AI, with a particular focus on how copyright protections continue to be safeguarded in the context of training data for generative AI models, as set out by the EU AI Act (2024) and the EU Copyright Directive (2019).

Our panel consisted of distinguished experts on AI, copyright and IP:

• Ania Helseth, EU Policy Manager at META
• Dessislava Fessenko, Attorney and Independent Expert on AI governance
• Pablo Asbo, Trilligent Advisory Board Member and expert on EU tech policy and IP
• Moderated by Timea Strihova, Senior Director and head of Trilligent Brussels and the UK

Key issues at the intersection of AI and copyright

With the unprecedented evolution of AI technology and the massive growth of generative AI models, several issues around intellectual property protections arise in our legal framework – both on the input side, so copyright protections of data being used for training AI systems and commercial exploitation, and on the output side the ownership and potential patent protection of content generated by AI.

On the input side, several major landmark court cases have made headlines globally, including the New York Times v. Microsoft/OpenAI (2023, US) and Getty Images v. Stability AI (2023, UK). In both cases, the claimants are seeking damages for the use of their copyrighted content for the training of the plaintiff’s AI models without permission, and the judgements may have the power to shape global copyright frameworks. On the output side, a good example is Thaler v. Comptroller-General of Patents (2021, UK), where Mr. Thaler sought to have his AI model recognized as an inventor, but the ruling judged only humans can own intellectual property. Another illustration of the question around non-human IP ownership is the famous monkey selfie case (no, the monkey cannot own the rights to its own selfie).

Generative AI is a rapidly evolving field, with AI systems thrive on collaboration between ecosystems, offering opportunities for everyone. However, questions remain on how to monetize AI effectively.  Observing the right balance between innovation and copyright protection is key in this debate, which needs to address both legal and ethical questions. Legal questions focus on protecting creators without undermining others in the ecosystem, while ethical issues arise when technology risks exploitation. Fair access to data underpins this balance, as differences in the amount and quality of data feeding into AI systems’ training can make or break the success of the AI model, leading to important questions about competition, as well as to unintended biases.

What does the regulatory framework look like globally, and in the EU?

From a regulatory standpoint, it is also important to note differences between different jurisdictions and their approach to copyright protections globally. Some countries, such as Japan and Singapore have more flexible copyright protections that allow for easier use of copyrighted materials for training AI model. US law foresees an exception of copyright protection in case of fair use (such as criticism, comment or research) and EU law allows for exceptions for scientific research and provides opt-outs from data scraping for rightsholders. Needless to say, this creates a maze of regulatory frameworks to navigate for providers of AI systems.

Looking at the EU specifically, copyright protections in the context of AI are governed by the recently adopted EU AI Act from 2024, which will be fully applicable in 2026, and by the EU Copyright Directive, or also known as the Digital Single Market – DSM Directive, from 2019, which saw significant delays in its transposal to national law that was originally foreseen for 2021.

The EU AI Act aims to address issues around copyright through transparency requirements which would require the disclosure of information about data used for training the AI model, allowing rightsholders to verify whether their work was used without permission. While this measure is a step in the right direction, the level of detail and format of these transparency reports remains to be seen, with important aspects to consider around undue burden for AI providers and trade secret protections in making information available, while ensuring that sufficient data is available to allow rightsholders to enforce their copyright protections.

On the other hand, the Copyright Directive provides certain exceptions to copyright protections, such as data mining for scientific purposes, as well as exceptions for commercial use if the data is obtained lawfully. The Copyright Directive also guarantees an opt-out option for rightsholders, should they not want their work to be scraped. However, a key issue in this context is enforceability and practical solutions to help guarantee these opt-out rights. On the output side, the Copyright Directive protects human creators only, so important questions remain around the nuances of human prompting and contribution to AI generated content, as well as terms of use relating to ownership of such content.

What are the solutions and how can you navigate the legal framework?

There seems to be consensus that, at the EU level, we have enough regulation and legislative tools (such as the transparency requirements and opt-outs) to avoid greyzones. However, the devil is in the detail and enforcement and practical implementation will be key to determine where balance in access lies, to ensure that rightsholders enjoy reasonable protections while AI providers have access to data necessary for driving innovation. The EU’s and Member States’ work in this regard will be crucial over the next couple of months.

At the same time, it is equally important to take step back and look at the AI landscape wholistically, from a global perspective, as most large AI providers span across multiple borders, and it can quickly become a nightmare to try coordinate multiple jurisdictions. Europe needs to continue to coordinate with international actors to ensure legislative coherence. Multilateral organisations, such as the World Intellectual Property Organisation – WIPO and the Organisation for Economic and Social Development – OECD, among others, have also been looking at the issue of AI, IP and copyright, and efforts are underway to facilitate the conversation at the global level.

At Trilligent, we follow this issue very closely, and our global team is well-positioned to help clients navigate the different regulatory conversations globally. If you are interested in further exploring the intersection of AI, IP and copyright, or would like to make sure your voice is heard in the debate, please feel free to reach out.