Government Draft Restricts BNetzA to Federal AI Market Surveillance
German government draft limits BNetzA to federal AI market surveillance, prompting debate over state oversight, clear enforcement and the AI Act’s innovation goals.
The German government’s latest draft proposal confines the Federal Network Agency (BNetzA) to market surveillance of AI systems at the federal level, a move that narrows the agency’s originally envisaged remit and raises questions about oversight across Germany’s federal states. The draft replaces an earlier version that would have empowered BNetzA to act as the central regulator for AI deployments in both federal and state jurisdictions. The change directly touches on commitments in the coalition agreement to implement the EU AI Act in a manner that preserves innovation while avoiding fragmented market supervision.
Details of the revised draft
The revised text limits BNetzA’s authority to federal administrative domains and federal public procurement, rather than extending its reach to state-run institutions and markets. That contrasts with the earlier reference draft, which proposed a more consolidated national supervisory model to cover both levels of government. Officials involved in drafting framed the revision as a clarification of competencies, but the narrower scope will transfer responsibility for state-level enforcement back to the Länder.
Background and coalition expectations
When CDU, CSU and SPD set out their coalition agreement, they explicitly sought an implementation of the AI Act that would be “innovation-friendly and low on bureaucracy,” and they stated that market surveillance should not be fragmented. The earlier government text reflected that ambition by centralizing oversight. The current draft, however, appears to balance centralization with deference to federalism, reigniting debate about how to reconcile the coalition’s efficiency goals with constitutional divisions of authority.
Implications for state regulators
If the draft becomes law in its present form, state regulatory bodies will retain or regain primary responsibility for supervising most AI applications within their territories. That outcome may increase administrative variation among Länder, as each state could develop different approaches to compliance monitoring and enforcement. Legal experts warn that divergent practices could complicate cross-border market surveillance and create legal uncertainty for companies operating in multiple states.
Industry and innovation concerns
Business groups and technology firms have voiced concern that a fragmented supervisory landscape could undermine the coalition’s promise to keep the AI Act implementation innovation-friendly. Companies said that a single national authority could offer clearer rules and streamlined compliance processes, whereas a patchwork of state-level regulators might mean duplicative audits, inconsistent interpretations and higher costs. At the same time, some regional stakeholders argue that state regulators can better account for local market conditions and specific public-sector uses of AI.
Enforcement logistics and coordination mechanisms
Central to the debate is how enforcement will operate in practice if states remain largely in charge. Policymakers are considering coordination mechanisms, data-sharing protocols and model procedures to harmonize oversight across jurisdictions. The government draft alludes to cooperative arrangements but stops short of prescribing a full central command, leaving the details of practical cooperation to subsequent regulation and intergovernmental agreements.
Germany now faces a choice between a tightly centralized supervisory model, which promises uniformity, and a federalized approach that may preserve regional autonomy but risks divergent enforcement. Advocates for centralization emphasize the benefits of scale and consistent interpretation of the AI Act, while proponents of state supervision stress subsidiarity and responsiveness to local needs.
The draft will proceed through interministerial review and discussions with the Länder, and parliamentary debates are expected to follow as the government prepares national legislation to transpose the EU AI Act. Stakeholders from industry, civil society and legal circles are likely to press for clarifications on jurisdiction, appeals procedures and the practical interface between federal and state enforcement.
Looking ahead, the resolution of this issue will determine how Germany balances regulatory clarity with federalist principles in the implementation of the AI Act. The final design of market surveillance will shape compliance burdens for businesses and the uniformity of protections for citizens, making the coming legislative negotiations a consequential moment for AI governance in Germany.
