Germany could lose €155 billion a year in extra economic value if access to AI training data gets tighter, according to a study released Tuesday by the Munich-based ifo Institute. Why it matters: the study puts a price on a policy choice Europe is actively debating — how strictly to limit the data used to train AI models — and argues that overly restrictive rules would weaken both the models themselves and the wider economy that uses them.
October 13, 2026
What did the ifo study find?
The institute estimates that generative AI could add about €440 billion a year to Germany’s economy by 2034. Under a scenario where training-data access becomes more restricted, that gain falls by €155 billion a year — roughly $173 billion at the exchange rate cited in the reporting.
Those are model-based projections, not measured results. They depend on assumptions about how quickly businesses adopt AI, how capable future models become, and how much data rules change in practice.
Where would the €155 billion loss come from?
ifo breaks the shortfall into four parts. The largest share, €75 billion, comes from weaker models. Slower adoption by businesses accounts for €45 billion, reduced research output for €25 billion, and less AI development inside Europe for €10 billion.
In other words, the study’s argument is not only that European labs would train less capable systems. It is that downstream effects — companies adopting AI more slowly and researchers producing less — would compound the direct hit.
What is ifo recommending?
The authors urge policymakers to keep the EU’s existing text-and-data-mining exemption for commercial uses, paired with proportionate transparency requirements and a workable opt-out standard for rights holders. They also say regulatory uncertainty is already discouraging investment in Europe.
ifo researcher Oliver Falck summarised the trade-off in the reporting as a choice: wanting a leading European AI model while also making training-data access unnecessarily complicated are goals that conflict with each other.
Why it matters
Europe is trying to do two things at once: protect creators and privacy, and build AI systems that can compete with those from the United States and China. This study is one side of that argument, quantified. It does not settle the copyright or consent questions, and other stakeholders — publishers, artists and data-protection advocates — make the opposite case that unlicensed training undermines their work.
For readers outside Germany, the number still matters. Rules set in the EU tend to shape how global AI companies handle training data everywhere, so a tighter European regime could affect which models are available, what they cost, and how they were trained — well beyond Europe’s borders. That is the “everything AI, tested” takeaway: data policy is not abstract. It shows up in the tools you can actually use.
Frequently asked questions
How much could generative AI add to Germany’s economy?
ifo estimates about €440 billion a year by 2034 under its baseline scenario.
How much would tighter data rules cost, according to the study?
€155 billion a year in lost additional value, split between weaker models, slower adoption, lower research output and reduced AI development in Europe.
Is this a forecast or a measured result?
It is a model calculation — a projection based on assumptions — not a measurement of something that has already happened.
Sources: Reuters