from the freedom-of-knowledge dept
This series of posts explores how we can rethink the intersection of AI, creativity, and policy. From examining outdated regulatory metaphors to questioning copyright norms and highlighting the risks of stifling innovation, each post addresses a different piece of the AI puzzle. Together, they advocate for a more balanced, forward-thinking approach that acknowledges the potential of technological evolution while safeguarding the rights of creators and ensuring AI’s development serves the broader interests of society. You can read the first, second, third, and fourth posts in the series.
In April 2007, Janez Potočnik, then European Commissioner for Science and Research, introduced the concept of the Fifth Freedom: the “freedom of knowledge.” His vision was simple but ambitious—enhance Europe’s ability to remain competitive through knowledge and innovation, the cornerstones of prosperity. Fast forward to today, the momentum for this Fifth Freedom is building once again, with both the Letta Report and the Mission Letter of the new EU Commissioner for Startups, Research, and Innovation emphasizing its significance.
But how does this freedom of knowledge intersect with creativity and copyright?
AI, Learning, and the Limits of Copyright
Machine learning (ML) systems learn in a way strikingly similar to humans—by observing and copying. This raises an important question: should ML systems be allowed to freely use copyrighted materials as part of their learning process? The answer is not just about technology; it goes to the heart of what copyright law aims to protect.
Traditionally, copyright protects the expression of ideas, not the ideas themselves. This is an important distinction because it allows others to take inspiration, innovate, and build upon ideas without infringing on someone else’s creative output. When an ML system is trained, it doesn’t care about specific creative choices—like the lighting or composition of a photo. It just wants to learn the underlying pattern, such as recognizing a stop sign. Similarly, a natural language model uses written text not because it appreciates the author’s unique writing style, but because it needs to learn the structure of language.
Humans also do this all the time. We often replicate expressions when learning, but our goal is not to plagiarize someone’s unique creative touch—it’s to grasp the idea behind it. This concept is embedded in many legal precedents. For instance, in the American Geophysical Union v. Texaco case, photocopying was used not for the beauty of the prose, but simply as a convenient way to access scientific ideas. Similar issues arise in cases about software interoperability, functional objects like clothing designs, and even in disputes over yoga routines. Copyright should protect creative expression—not the ideas, facts, or functional elements that underpin them.
Why This Matters for Machine Learning
This distinction is particularly important for ML. If we allow copyright law to get in the way of machines learning from data for purely non-expressive purposes, we’re potentially hampering technological advancement. Allowing ML systems to copy for learning—without trying to replicate the creative aspects of the original work—is essential for innovation. This is not just a matter of advancing technology but also of staying true to the spirit of copyright law, which is meant to balance the interests of creators and the public good.
However, as Professor Lemley has pointed out from a U.S. law perspective, the freedom for ML to learn should have limits. If an ML system is being trained to create a song that mimics the style of Ariana Grande, it’s no longer just about learning—it’s about copying a creative expression. In such cases, the question of whether it qualifies as fair use becomes much tougher. Yet, even here, it’s crucial that copyright doesn’t end up controlling unprotectable elements like a musical genre or a broad artistic style.
Finding the Balance: Innovation and Protection
The concept of the Fifth Freedom—freedom of knowledge—cannot thrive if copyright is used to restrict learning and innovation. We need a balanced approach: one that protects the hard work of creators, while ensuring that copyright doesn’t stifle the fundamental right to learn, innovate, and build upon existing knowledge. This is especially relevant now, as AI and machine learning shape the future of creativity and the knowledge economy in Europe. If we get this balance right, we can ensure that both creativity and innovation continue to flourish in the digital age.
Caroline De Cock is a communications and policy expert, author, and entrepreneur. She serves as Managing Director of N-square Consulting and Square-up Agency, and Head of Research at Information Labs. Caroline specializes in digital rights, policy advocacy, and strategic innovation, driven by her commitment to fostering global connectivity and positive change.
Filed Under: access to knowledge, ai, copyright, creativity and ai, fifth freedom, freedom of knowledge, knowledge, machine learning
