According to a model calculation by the Ifo Institute, Germany risks losing a substantial part of its value creation potential through Artificial Intelligence (AI) if access to training data for AI models continues to be restricted. Generative AI could potentially add 440 billion euros to the country’s annual economic output by 2034. However, this figure would fall by 155 billion euros if the access of AI developers to necessary training data becomes difficult.
The projected potential of 330 billion euros is composed of broad AI usage, 65 billion euros from more productive research, and 45 billion euros stemming from AI development within Germany.
Oliver Falck, head of the Ifo Centre for Innovation Economics and Digital Transformation, and his co-author Martin Thelle from Implement Consulting Group, highlighted that stricter regulations could negatively impact the economy through four specific channels: slower dissemination (costing 45 billion euros), less powerful models (costing 75 billion euros), lost research potential (25 billion euros), and reduced developmental capacity across Europe (10 billion euros).
Powerful AI language models require massive and diverse data sources. Currently, the largest copyright-free dataset in Europe comprises only two to three trillion tokens, whereas leading international models utilize between 15 and 36 trillion tokens. Consequently, the authors are advocating for maintaining the text- and data-mining exception within the EU Copyright Directive to reduce uncertainty for investors.
