Germany Hands AI Oversight to Telecom Regulator as Munich Court Sets Liability Precedent
14.06.2026 - 04:53:25 | boerse-global.de
A Munich court has ruled that companies are fully liable for AI-generated summaries in search results, treating them as their own content rather than mere links to third-party information. The late-May decision by the Landgericht München I ordered Google to stop distributing false factual claims about two publishers and forced the company to cover 80 percent of the legal costs for now — a ruling that is not yet final.
The judgment landed just weeks before Germany’s parliament passed the AI Market Surveillance and Innovation Promotion Act on June 12, a law that redesigns the Bundesnetzagentur as the country’s central AI supervisory authority. The legislation aims to implement European Union requirements while leaving room for innovation.
Under the new framework, the telecom and energy regulator will establish a coordination and competence center and operate what it calls an AI real-world laboratory, where businesses can test new applications under official supervision. A first evaluation of the system is due after 18 months, a second after three years.
The price tag is substantial: the federal government faces one-time costs of €4 million plus recurring annual expenses of €15.9 million, while Germany’s states shoulder roughly €33.1 million each year. Violations of the rules can draw fines of up to €50,000.
The governing coalition of CDU/CSU and SPD voted in favor of the bill. The opposition parties AfD, Greens, and the Left voted against, with critics specifically targeting the creation of 43 new permanent posts at the agency and what they called insufficient protection of fundamental rights. Motions from the AfD aimed at strengthening German AI sovereignty were referred to parliamentary committees.
Data protection authorities are meanwhile urging caution. The state commissioner for data protection in Lower Saxony warned that employees are using private AI accounts for professional tasks, a practice dubbed “shadow AI” that risks data breaches, security gaps, and the leakage of trade secrets. Under the EU AI Act, companies must ensure their staff have adequate AI competence. Experts recommend clear internal directives, regular training, and official privacy-compliant alternatives. A study from early this year underscores the urgency: more than half of all AI agents currently operate without any monitoring, and 49 percent of executives cite a lack of internal coordination as the biggest obstacle.
Starting August 2, the European Commission’s code of conduct on labeling AI-generated content takes effect, imposing transparency obligations particularly for deepfakes and AI-written texts on public-interest topics.
Separately, the Federal Commissioner for Data Protection has raised objections to the planned annual tax law for 2026. Finance authorities intend to train AI models using real, non-anonymized tax data. The BfDI warned that personal data could become permanently embedded in the models and demanded statutory deletion rules and technical safeguards to minimize data use. His deadline for submitting objections was June 12, the same day the new AI oversight law passed.
