Brando Benifei and Axel Voss—names tied to the AI Act’s legislative path—sent a letter, per MLex around September 17, 2026, warning that the European Data Protection Supervisor (EDPS) lacks technical, financial, and staffing resources to oversee AI used by EU institutions, including high-risk systems and a regulatory sandbox.
The EDPS, in budget documents and its 2026–2027 compass, identifies dozens of new AI Act tasks—public figures cite about 54—and notes that an initial allotment of just two full-time equivalents sufficed only for preliminary preparation. Market-surveillance and notified-body roles demand another scale of engineering, legal capacity, and audit muscle.
The MEPs ask the Commission to allow access to Digital Europe programme funds to reinforce that capacity. Without that budget bridge, the risk is a cruel asymmetry: ambitious rules for European public AI and a supervisor without a lab or headcount to enforce them inside the institutions themselves.
The problem is not headcount alone. Infrastructure to evaluate models, sandbox protocols, talent that understands both GDPR and AI Act high-risk taxonomies, and incident response capacity are required. Two FTEs do not build that; they barely draft roadmaps.
Politically, the letter arrives as Brussels touts global normative leadership. An underfunded EDPS turns that leadership into facade: exporting the standard while the home inspectorate lacks staff. Benifei and Voss, with different histories but digital weight, seek to avoid that embarrassment before an incident erupts.
Member states watch with a double lens. Those wary of more EU spending will see another budget ask; those fearing opaque Commission AI in borders, asylum, or public procurement will see underfunding as a fundamental-rights risk. Parliament can use the letter in MFF talks.
For vendors selling to European institutions, the signal is enforcement uncertainty: slow sandboxes, late guidance, shallow audits. Paradoxically, that can favor incumbents who already comply “on paper” and hurt SMEs that need a real sandbox to iterate.
The EDPS is not the AI Office nor national market-surveillance authorities for the private sector; the focus here is AI use by the EU administration. Confusing mandates dilutes the funding ask. Benifei and Voss target the concrete institutional hole, not a rewrite of the entire AI Act org chart.
Neural Pulse reads the letter as a credibility stress test: without money for the EDPS, the AI Act inside the institutions risks becoming symbolic law. Opening Digital Europe to the supervisor will not alone fix governance, but it keeps oversight from being born already short-staffed.
Brussels journalists should track the Commission’s formal reply: timelines, budget lines, and promised FTEs. Without numbers, political empathy does not pay evaluation engineers. September 2026 marks the visible start of that fight, not its end.
Privacy civil society can ally with the signatories: a strong EDPS protects residents when European bureaucracy automates decisions. A weak EDPS leaves the AI Act as a poster on a Commission wall. In that sense, the Benifei–Voss letter is a request for operational oxygen.
Comparing with national agencies already hiring AI teams shows the relative lag at the institutional EU level. If Berlin or Paris audit more than the EDPS, the “Brussels leads” narrative becomes hard to sustain at 2027 international conferences.
The next board square is the autumn budget dialogue. If Digital Europe does not open, ad hoc reinforcements or internal reshuffles remain—always more fragile than a clear multi-year line for the 54 tasks the EDPS itself enumerates.
The institutional regulatory sandbox, if born without engineers, becomes a bureaucratic queue where Commission pilot projects wait for months. That discourages legitimate experimentation and pushes internal teams to deploy “in the gray” without oversight. Funding the EDPS is, on that reading, also an investment in safe public innovation—not only in brakes.
European public-sector unions should join in: HR systems, machine translation, and citizen-mail triage already use AI. A supervisor without capacity leaves workers without a technical interlocutor when they demand explainability. The Benifei–Voss letter opens that political door; bottom-up pressure must follow.
