EU AI Office issues first formal requests to model providers

The European Commission’s AI Office has sent its first formal information requests to dozens of general‑purpose AI model providers, the office said, an initial enforcement step under the EU AI Act aimed at probing model security, independent external evaluations and post‑market monitoring.

The requests were issued on Sept. 1, 2026, the AI Office told EU Perspectives, targeting developers of so‑called frontier models as part of a wider verification exercise using powers granted by the Act. The letters ask for technical documentation and evidence that providers have implemented safeguards, arranged for external conformity assessments where required, and set up systems to monitor models after deployment.

The move represents the first time the AI Office has used information‑gathering powers created by the EU AI Act to seek detailed operational data from general‑purpose model suppliers. Regulators are focusing on a trio of compliance priorities: how models are secured against misuse, whether independent evaluators have been engaged to test high‑risk behaviours, and how firms will conduct ongoing post‑market surveillance to detect harms once models are in use.

Commission officials described the step as a targeted, evidence‑gathering measure rather than an enforcement sanction. The AI Office said it will assess the responses to determine whether formal corrective or enforcement actions are necessary under the Act’s supervisory framework.

The action follows months of preparatory work by Brussels to define which models fall into the frontier or general‑purpose categories and to craft questionnaires that probe opaque aspects of model development. Industry lobbying has pressed for clarity on thresholds that trigger obligations such as third‑party conformity assessments and stricter post‑market obligations; the AI Office’s information requests appear intended to establish a factual baseline before setting enforcement priorities.

For HR teams and employers that licence or embed large language models and multimodal systems into recruitment, screening and employee‑facing applications, the requests signal potential new flows of vendor information. If the AI Office’s inquiries force providers to produce independent evaluation reports or share post‑market monitoring procedures, organisations buying those models could gain greater visibility into safety testing, mitigation of discriminatory outcomes and incident‑response commitments — details that affect procurement, supplier risk reviews and contractual warranties.

What the AI Office has not disclosed is which providers received requests, the specific deadlines for supplying materials, and how the office will handle sensitive intellectual property or trade secrets contained in technical dossiers. The EU Perspectives account notes the enquiries are being sent to “dozens” of firms but stops short of naming recipients or setting out the evidence standard the regulator will apply when judging compliance. The regulator also has not published a template of the information request or a redaction regime for commercially confidential material.

The lack of public detail leaves employers and HR leaders uncertain about how quickly they might see more detailed model‑level evidence from vendors, and whether vendors will be able to meet buyers’ demands for independent audit reports without breaching confidentiality or competitive constraints. Vendors that must produce third‑party evaluations could face longer procurement lead times and new negotiation points in supplier contracts.

Regulatory pressure on model transparency is likely to intensify across Europe as the AI Office processes responses and decides whether to open compliance proceedings. For HR leaders, that means procurement checklists and vendor assurances around bias testing, incident reporting and post‑market monitoring could become not just best practice but standard expectations in supplier contracts. The first information requests mark the start of a period in which both providers and buyers will learn how the EU AI Act’s requirements translate into the operational documentation and governance many workplaces rely on.