Germany could forgo €155 billion in additional annual economic value if access to data for AI training becomes more restrictive, the ifo institute said on 13 October, Reuters reported. The figure is an estimate from ifo’s model, rather than the measured cost of a rule already in force. The institute urged EU policymakers to retain the commercial text-and-data-mining exemption while setting proportionate transparency requirements and a workable standard for opting out.
Key points
- Ifo estimates generative AI could add €440 billion a year to Germany’s economy by 2034.
- In its restrictive-data scenario, the institute puts the reduction in that potential gain at €155 billion annually.
- Its model assigns €75 billion of the potential loss to weaker AI models and €45 billion to slower adoption.
- The authors favour retaining the EU exemption for commercial text and data mining, with transparency and opt-out provisions.
Ifo’s €155 billion estimate rests on four effects
Ifo estimates that generative AI could add €440 billion a year to Germany’s economy by 2034. Its calculation puts the annual gain €155 billion lower if access to training data is restricted. Both figures concern potential economic value, and the smaller figure is the amount ifo attributes to tighter access in that scenario. They are not estimates of what Germany currently earns from AI.
Ifo’s model assigns €75 billion of the estimated €155 billion potential loss to weaker models. The institute assigns another €45 billion to slower adoption, separating a possible effect on the systems available to users from an effect on how widely those systems are taken up. That distinction matters to its policy case: access to training material is presented as affecting both the development of AI and its use across the economy.
Ifo assigns €25 billion of the estimated loss to lower research output and €10 billion to reduced AI development in Europe. The latter category extends beyond German use of AI to the location of development work. Taken together, the four amounts are the components of ifo’s model calculation, not separately observed losses. The institute’s argument therefore depends on how strongly restrictions on training data would affect each of those activities.
European policymakers and technology companies are weighing how to manage AI risks without slowing innovation while the region seeks to catch up with the United States and China in advanced models, Reuters reported. Ifo’s estimate gives that debate a German economic figure. It also places model quality ahead of adoption as the larger of the two effects in its calculation, making the conditions for training a central part of its case.
Ifo backs the EU text-and-data-mining exemption
The authors urged policymakers to preserve the EU’s text-and-data-mining exemption for commercial uses. They paired that request with proportionate transparency requirements and a practical opt-out standard. Their position does not ask policymakers to abandon oversight of training data. It asks them to retain a route for commercial use while defining how rightsholders can object and what information users of the exemption should provide.
Oliver Falck, an ifo expert, put the institute’s preference in terms of Europe’s ambitions for its own models: “Anyone who wants a leading European AI model cannot at the same time make access to training data unnecessarily complicated. You cannot have both.” That is Falck’s policy argument, alongside an economic estimate whose size depends on the modelled effects of restricted access.
The authors also said regulatory uncertainty was already deterring investment in Europe. That claim gives their proposed opt-out and transparency standards a second purpose beyond their estimate of future economic value: setting terms on which training data can be used. Ifo has separately criticised the German government over its spending of debt-funded fiscal room on purposes other than the infrastructure and defence for which it was intended.
