Tighter AI data access could cost Germany €155 billion a year, ifo says

By Cian Muenster and Leo Marchandon

Oct 13 (Reuters) - Germany risks losing €155 billion ($173.43 billion) in additional annual value added if access to AI training data is tightened, the ifo economic institute said on Tuesday.

The findings come as European policymakers and technology companies seek to balance the risks posed by AI with the need to foster innovation, as the region races to catch up with the United States and China in developing advanced 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," ifo expert Oliver Falck said.

Generative AI could add €440 billion a year to Germany's economy by 2034, but the loss would stem from slower AI adoption (€45 billion), weaker models (€75 billion), lower research output (€25 billion) and reduced AI development in Europe (€10 billion), according to the influential German institute's model calculation.

"Access to diverse, high-quality training data is essential for competitive AI. Unnecessary restrictions and legal uncertainty make its development harder," German AI lab Aleph Alpha CEO Ilhan Scheer told Reuters.

"Europe should make full use of its existing legal framework," he added.

The authors urged policymakers to preserve the EU's text-and-data-mining exemption for commercial use, while setting proportionate transparency requirements and a practical opt-out standard. They said regulatory uncertainty was already deterring investment in Europe.

AI can only realise its potential if access to data is not unnecessarily restricted, said Florian Schewor, DACH managing director at data processing company Celonis, adding that German and European industries hold vast amounts of largely untapped data.

The Munich-based institute recently criticised the German government for using debt-financed fiscal leeway intended for infrastructure and defence spending for other purposes.

($1 = 0.8937 euros)

(Reporting by Cian Muenster and Leo Marchandon in Gdansk, editing by Izabela Niemiec and Anil D'Silva)