EU AI Act requirements for AI hiring tools — Commission probes

The European Commission’s AI Office has sent formal information requests to more than 30 artificial‑intelligence providers, a move that tightens scrutiny relevant to EU AI Act requirements for AI hiring tools.

This week the newly empowered office began asking for documentation and evidence on model safety, independent evaluations, post‑market monitoring and copyright or transparency measures. Agence Europe reported that recipients include major model-makers such as OpenAI, Anthropic and Google, among others.

The requests are a preliminary enforcement step under the AI Act designed to gather the technical dossiers regulators will use to decide whether to open investigations or require corrective measures. Actuia described the outreach as part of a broader phase in which legal texts under the AI Act are being turned into operational oversight, with the Commission signalling it will move from rule‑making to active compliance checks.

For employers and HR vendors that embed general‑purpose models into hiring, screening, employee surveillance or people‑analytics, the letters could bring downstream effects. If providers are required to produce independent conformity assessments, post‑market monitoring records or incident reports, HR platforms that rely on those models may face new documentation demands and contractual change requests to meet customer or regulator expectations.

The Commission’s stated focus areas mirror the AI Act’s key compliance pillars: demonstrable safety controls for high‑risk applications, third‑party testing, ongoing monitoring of deployed systems and clarity about copyrighted training data and model transparency. The information drive is intended to let EU authorities determine whether providers’ current processes satisfy the Act’s obligations for models that are used, or reasonably foreseeably used, in high‑risk contexts.

Regulatory scrutiny of foundational models has intensified across jurisdictions; national data protection authorities and competition regulators have already signalled parallel lines of inquiry into training data, model explainability and consumer‑facing safety controls. The Commission’s step — taking direct requests to multiple providers at once — reflects an enforcement posture that could shorten the lag between rule adoption and practical obligations for downstream vendors and end users.

What wasn’t disclosed in Agence Europe’s reporting is which specific models or deployments the letters target, the exact content of the Commission’s information requests, and the timetable for providers to respond. Neither the Commission nor the firms named in press reports have published the full scope of documents sought, whether certain kinds of internal test results or licensing records for training data will be treated as mandatory, or what thresholds would trigger an investigatory escalation.

For HR teams and vendors, that lack of detail creates immediate contract and compliance ambiguity. Many people‑analytics and recruitment platforms contractually incorporate third‑party model warranties and audit clauses; those agreements may need renegotiation if providers are compelled to alter post‑market monitoring, incident‑reporting practices or usage restrictions. Legal teams should be watching for formal guidance from the Commission on data to be supplied and any sector‑specific interpretations that clarify when a particular hiring workflow counts as a high‑risk AI system under the Act.

The Commission’s information requests mark an early enforcement rhythm under the AI Act that will test how rapidly European oversight can translate into obligations for AI suppliers and the businesses that depend on them. For HR leaders, the immediate task is operational: map where general‑purpose models feed into hiring and workplace analytics, catalogue contractual dependencies, and be prepared to seek evidence from vendors that their models, and the documentation underlying them, will meet the new EU standards.