The EU Data Mining Manifesto

For a Sovereign, Competitive, and Intelligent Europe

July 24, 2026

PREAMBLE

Europe is the cradle of modern scientific inquiry. From the printing press to the World Wide Web, our continent has championed the creation and dissemination of knowledge as the engine of progress. Today, we stand at the threshold of a new era defined by Artificial Intelligence. This term denotes a wide range of technologies that combined are changing the way knowledge is created, discovered, synthesized, and multiplied. While this process is still in its infancy, the transformative potential of these technologies for our societies and economies is clear and unequivocal.

However, on the cusp of this revolution, Europe risks becoming a museum of past glories rather than a hotbed for future breakthroughs. While other global powers treat AI as a strategic priority and pave the way for its development with flexible data usage frameworks, the European innovation environment is increasingly plagued by legal uncertainty, creating legal risks and disincentivising research and development of frontier technology.

The copyright framework governing text and data mining (TDM) counts highly among numerous issues that need to be addressed if Europe is to retain a position in the global competition for AI discovery. Europe must secure the legal environment where AI creation is not just possible, but is world-leading. If Europe does not train its own AI, it will depend on AI trained elsewhere — imbued with values, languages, and interests that are not our own. To protect European language, culture, and truth, we must be free to train on European data without prohibitive transaction costs or legal risks. Frontier models are a strategic asset and Europe must ensure it has its own. Building AI gigafactories makes no sense if the legal system will prevent their full utilisation for lack of training data.

OUR DECLARATION

We, the undersigned researchers, entrepreneurs, and industry leaders wish to declare our utmost support for a copyright framework that facilitates the creation of frontier AI models here in Europe.

This framework should be based on the following principles:

1. Gaining knowledge from a work is not infringement of this work

We reject the notion that analysing lawfully accessible data could be an infringement of intellectual property — lawful access to a work should also guarantee the right to mine and derive the information it contains. Training AI models is not copying content — models only store numerical values (weights) that parametrise probability distributions, capturing statistical patterns learned from the training data, not the works themselves. Copyright law was never intended to hinder the ability to create new works, which is what AI models are built to do. To restrict, or block this process is to restrict learning itself. We assert that the act of observation is not copyright infringement and no part of training AI models should be restricted on copyright grounds.

2. Technical Solutions Must Serve the Law, Not Distort It

The emergence of workable, machine-readable opt-out mechanisms and other technical AI standards is a legitimate and desirable goal. However, technical standards must reflect legal reality, not reshape it. Any adopted protocols must be neutral, interoperable, and developed through open, consensus-driven multistakeholder processes, and not selected for political favours or commercial interests. A technical framework that is rushed, captured, or unworkable risks paralysing the ecosystem it was meant to serve and undermining the delicate legal balance between private interests of rightholders and the rights of society at large.

3. The Unity of Research and Enterprise

The distinction between "non-profit scientific research" and "commercial innovation" is an artifact of the past. Today''s PhD student should become tomorrow''s founder. We advocate for a unified ecosystem where knowledge flows freely from the lab to the market. We cannot shut our startups out of access to research results and data and expect them to become global champions. The gap between university research and the market has for years been Europe''s Achilles heel. We must strive to bridge this gap, by removing the legal distinction between scientific and commercial data mining.

CALL TO ACTION

We call upon the European Union institutions and its Member States to both uphold and fully and correctly harmonise the TDM and scientific exceptions, unify the standard for scientific and commercial TDM, standardise technical opt-outs to prevent abuse, refrain from taking any action to the contrary and provide the legal certainty required to keep our brightest minds working for the common benefit of this continent.

The future is written in data. We cannot afford to be left behind.