Brussels, BELGIUM – Restricting the EU’s current text-and-data-mining (TDM) framework – the copyright rules that allow AI models to be trained in Europe today – could cost the EU economy up to €600 billion annually.
This warning comes from a landmark economic study launched at the European AI Roundtable on Copyright, hosted by the Computer & Communications Industry Association (CCIA Europe).
The findings arrive at a highly charged political moment in Brussels, as the European Commission recently launched a call for evidence to potentially reopen the 2019 Copyright Directive, paving the way for a new legislative proposal in early 2027.
This push to alter the rules enabling AI training comes just months before the EU AI Act’s copyright provisions take effect in August, creating serious market uncertainty and threatening the legal foundations that tech innovators rely on.
The independent study ‘The TDM Equation’ – conducted by Implement Consulting Group in collaboration with the Ifo Institute – puts a clear price tag on changing Europe’s copyright rules. While it finds that generative AI could generate €1.65 trillion in annual value for the EU, more than a third of that opportunity is at risk if the current TDM framework is restricted.
The study warns that limiting data access would trigger severe economic shocks across the EU. It could stall AI adoption through legal uncertainty, drive AI development and talent to more permissive foreign jurisdictions, and degrade AI model capabilities by up to 50% – directly harming high-value sectors such as healthcare, finance, pharma, and manufacturing.
To safeguard Europe’s digital competitiveness, the report sets out three recommendations for EU policymakers to preserve the framework that enables AI training today. The first is to maintain the commercial TDM exception, giving AI developers the legal certainty to scale.
The second is to keep ‘opt-out’ mechanisms workable by endorsing globally recognised machine-readable standards such as robots.txt, rather than unproven alternatives. The third is to avoid rigid mandatory licensing frameworks that create transactional bottlenecks, and instead allow voluntary commercial data-sharing partnerships to continue developing.
The following can be attributed to CCIA Europe’s AI Policy Lead, Boniface de Champris:
“The Commission should not reopen the rulebook just as the AI Act’s copyright provisions are about to take effect. Doing so would send the wrong signal to the market and suggest the EU does not trust its own rules. The current text-and-data-mining exception in the Copyright Directive is the cornerstone of training AI models in Europe, and it also enables homegrown AI champions to compete globally.”
“If the Commission mandates rigid licensing frameworks or unworkable opt-outs for AI training, it will add unnecessary friction, price European innovators out of the market, and undermine the EU’s ambition to lead the global AI race.”
The following can be attributed to Martin H. Thelle, Senior Partner at Implement Consulting Group:
“Europe has a large AI opportunity, but it will not be realised automatically. If policymakers make it harder to train, fine-tune, and deploy competitive AI systems in Europe, the region risks slower adoption, weaker capabilities, and less innovation. The challenge is to protect rightsholders in a way that remains workable in practice and does not undermine Europe’s wider competitiveness.”
Notes for editors
The independent study ‘The TDM Equation’ measures how restricting the EU’s current text-and-data-mining (TDM) framework would harm the European economy, with a more detailed breakdown of the four specific financial risks provided below.
⮞ A single year of delay in AI adoption across the EU, triggered by regulatory uncertainty, would cost the economy €175 billion. Currently, 53% of EU firms already cite unclear legal consequences as a primary barrier to adopting AI technologies.
⮞ Restricting data access would slash local European AI development by €60 billion annually, driving top-tier tech talent, developers, and venture capital away from the EU to more permissive foreign jurisdictions.
⮞ Restricting high-quality training data results in a 50% drop in complex reasoning performance for AI models. This directly threatens €275 billion in value across high-stakes professional sectors like healthcare, finance, and legal services.
⮞ Weaker, data-restricted models would severely stall innovation and research efficiency in highly knowledge-intensive industries, putting an additional €90 billion per year at stake for sectors like pharmaceuticals and advanced manufacturing.
