Brussels is being warned that fiddling with Europe’s AI training rules could torch €600 billion a year.
A new economic study says restricting the EU’s current text-and-data-mining framework, the copyright rules that allow AI models to be trained in Europe, could clobber the bloc’s economy.
The warning landed at the European AI Roundtable on Copyright, hosted by the Computer & Communications Industry Association Europe. It arrives as the European Commission has launched a call for evidence that could reopen the 2019 Copyright Directive and lead to fresh legislation in early 2027.
The timing is a bit of a mess. The push comes just months before the EU AI Act’s copyright provisions take effect in August, leaving tech firms wondering whether Brussels trusts its own paperwork.
The independent study, The TDM Equation, was conducted by Implement Consulting Group with the Ifo Institute. It claims generative AI could create €1.65 trillion in annual value for the EU, but more than a third of that could be at risk if the current TDM framework is tightened.
The study warns that limiting data access would send a nasty shock through the EU economy. It says legal uncertainty could stall AI adoption, push AI development and talent to friendlier jurisdictions and degrade model capabilities by up to 50 per cent.
That would hit high-value sectors including healthcare, finance, pharma and manufacturing, which are not exactly short of regulatory grief already. The report urges EU policymakers to preserve the framework that enables AI training today. It wants the commercial TDM exception kept in place so developers have enough legal certainty to scale.
It says opt-out mechanisms should stay workable through globally recognised machine-readable standards such as robots.txt, rather than untested alternatives. It warns against rigid mandatory licensing frameworks that could jam the system with transactional bottlenecks. Instead, it wants voluntary commercial data-sharing partnerships to keep developing.
CCIA Europe, AI policy lead, Boniface de Champris said: “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,” de Champris said.
Implement Consulting Group, senior partner, Martin H Thelle said: “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.”
The study claims a single year of delayed AI adoption, caused by regulatory uncertainty, would cost the EU economy €175 billion. It says 53 per cent of EU firms already cite unclear legal consequences as a primary barrier to adopting AI technologies.
Restricting data access would cut local European AI development by €60 billion a year, the study says. That could shift top-tier tech talent, developers and venture capital towards more permissive foreign jurisdictions.
The report says restricting high-quality training data causes a 50 per cent drop in complex reasoning performance for AI models. That would threaten €275 billion in value across healthcare, finance and legal services.
Weaker, data-starved models would slow innovation and research efficiency in knowledge-intensive industries, putting another €90 billion a year at risk in pharmaceuticals and advanced manufacturing.
