Open Source AI Gains Momentum in Europe, Navigating Copyright and Regulation
The European Union is fostering a unique environment for open source AI development, leveraging recent legislation to incentivize innovation and potentially challenge the dominance of US and Chinese tech giants. This approach centers on a concept known as “public AI,” systems built for the public good rather than pure profit, and is underpinned by both the EU’s Artificial Intelligence Act and the 2019 Copyright in the Digital Single Market Directive (CDSM).
For a quarter of a century, open source software has been a disruptive force, evolving from a niche movement to the backbone of modern computing. As one author noted decades ago, open source offered a powerful alternative to traditional, proprietary software models. Today, it powers everything from the internet to mobile phones – Android being a prime example – and dominates the world of cloud computing. Remarkably, for the last ten years, nearly 100% of the world’s top 500 supercomputers have run on the open source operating system Linux.
The rise of generative AI is now extending this trend, but with new complexities. The Open Source Initiative has begun defining what “open source” truly means in the context of AI, which extends beyond just lines of code. The EU is taking a proactive stance, explicitly recognizing the importance of open source in the field of AI. Notably, open source AI systems are exempt from stringent documentation requirements outlined in the EU’s recent Artificial Intelligence Act.
This exemption is a significant incentive for AI developers within the EU, particularly for academic researchers already operating on limited budgets. Further bolstering this trend is the 2019 CDSM Directive, which allows research institutions to text and data-mine “works or other subject matter to which they have lawful access” for scientific research. This process, encompassing “any automated analytical technique aimed at analyzing text and data,” directly supports the development of AI models.
According to analysis by Paul Keller, Article 3 of the CDSM Directive effectively covers the entire AI model development process – from data acquisition to publication – for non-commercial projects. Specifically, if a model is released under an open-source license and is not commercialized, the reproductions and extractions made during training are legally protected. This creates a pathway for researchers to freely develop and share AI models without facing copyright challenges.
This confluence of legislation has given rise to “public AI” systems – AI built by organizations prioritizing public benefit over profit. These systems are defined as those focused on “creating public value rather than extracting as much value from the information commons as possible.” Their importance lies in offering an alternative to the profit-driven approach of major AI companies, particularly as concerns grow about a potential “AI bubble.”
Moreover, public AI presents a strategic opportunity for the EU to compete with the US and China. Acknowledging that Europe likely cannot match the venture capital investment of the US or the state-backed support enjoyed by Chinese firms, the EU is forging its own path. By embracing fully open source models and leveraging the CDSM Directive’s research provisions, Europe can offer a distinct approach to generative AI – one that may address concerns about the ethical development and deployment of these powerful technologies.
This uniquely European approach to generative AI, built on open source principles and supported by forward-thinking legislation, could prove to be a crucial differentiator in the rapidly evolving landscape of artificial intelligence.
