Your AI Screening Tool Is Probably Illegal in Europe. Here's What the EU AI Act Actually Requires.
The EU AI Act classifies AI-powered recruitment screening as high-risk — and most companies using these tools today are not compliant. This post breaks down what Articles 13, 14, and 15 actually require, why 'our vendor said it's compliant' is not a legal defence, and how voice-first AI changes the compliance picture.

The EU AI Act has officially classified AI-powered recruitment screening as high-risk — placing it in the same regulatory category as medical devices and critical infrastructure. Most hiring teams using AI to screen candidates today are not compliant. Most of them don't know it yet.
What "High-Risk" Actually Means in Practice
High-risk classification isn't primarily about banning AI from hiring decisions. It's about what the AI must be able to prove about itself before it can legally be deployed.
Article 13 of the EU AI Act requires high-risk AI systems to provide sufficient transparency for operators to understand the system's outputs. Article 14 requires that humans can effectively oversee the system — not just technically review decisions, but meaningfully intervene. Article 15 mandates that systems maintain logs enabling audits of every decision made.
If your AI screening tool scores a candidate 72 out of 100 and you don't have a breakdown of which factors produced that number, your legal exposure starts right there. If a rejected candidate requests an explanation — something EU residents are entitled to under both the AI Act and GDPR's Article 22 — and your vendor can't provide it, you are the one on the hook. The law assigns liability to the deployer. Not the vendor.
The "We're Just Using a Vendor" Defence Doesn't Hold
The EU AI Act distinguishes between providers — companies that build and sell AI systems — and deployers — companies that integrate them into their operations. Both carry legal obligations. Deployers must conduct conformity assessments, maintain technical documentation, and implement human oversight mechanisms that go beyond clicking a reviewed button in a dashboard.
"We purchased a tool our vendor said was compliant" is not a conformity assessment. It is a purchase order.
Vendors operating in Europe are issuing compliance statements — often vague, often designed more to reassure customers than to transfer legal coverage. That statement documents what you were told. It doesn't transfer what you owe. If your tool screens 400 candidates over a six-month campaign and you cannot produce audit trails showing how decisions were reached, you have an exposure that no compliance FAQ resolves.
An estimated 70% of European companies using AI in their hiring processes have not completed a formal conformity assessment [1]. That number will shrink quickly once the first enforcement actions begin — and HR technology, where AI deployment is wide and documentation is thin, is a predictable place to look.
The Good News: Compliant Design Is Also Better Hiring
Here is the part worth pausing on. The AI Act's requirements — explainability, human oversight, auditable decision logs — are not bureaucratic friction. They are a description of how a well-designed AI recruitment system should already operate.
A screening tool that cannot explain why it ranked a candidate highly is not giving you real signal. It is giving you a number. Numbers without provenance teach hiring managers to either distrust the AI and override it arbitrarily, or to defer to it blindly — neither produces better hires than what you had before. A tool that produces a structured conversation record for every candidate, with qualification reasoning your recruiters can actually read, gives you something the old process never offered: an evidentiary foundation for every decision. Useful the day you receive a discrimination complaint. Equally useful the day you want to know whether your screening criteria are actually predicting who becomes a strong hire.
The companies building for compliance are accidentally building better products.
How Asendia AI Approaches This
Asendia AI is a voice-first AI recruiter that conducts live screening conversations — not a scoring algorithm processing resumes against keyword lists. That distinction matters for EU AI Act compliance in two concrete ways.
First, the audit trail is a transcript, not a black-box score. Every conversation is a real, documented interaction. When a candidate asks why they progressed or were passed over, the answer exists in the record: here is what was discussed, here are the criteria that were evaluated, here is the qualification summary your recruiter received. That is Article 13 transparency by design, not retrofit.
Second, Asendia is built around human-in-the-loop oversight. Recruiters receive ranked shortlists with structured qualification notes and verbatim conversation excerpts. The AI handles every first screening call — 24 hours a day, 7 days a week, within hours of each application, without adding headcount — but humans make every subsequent decision. That is the oversight architecture Article 14 is describing.
Asendia connects directly to your existing ATS, so screened candidates arrive in your normal pipeline with a complete conversation record attached. Recruiting agencies use it to absorb application volume without staff additions, with a compliance-ready paper trail that requires no additional compliance workflow. For teams thinking through where AI recruitment tools genuinely move the needle versus where they add a layer of false confidence, the post on agentic recruiting covers why that distinction compounds over time.
Final Word
The EU AI Act is not a European problem you can manage from outside Europe. Any organisation deploying AI that touches EU residents — in hiring, screening, or assessment — is within its reach, regardless of where that organisation is headquartered. The enforcement timeline is accelerating, and HR technology is an obvious early target: deployment is extensive, documentation is sparse, and the liability exposure is straightforward to establish. The practical path forward is not to remove AI from your recruitment process. It is to replace opaque scoring systems with tools that produce explainable outputs, maintain conversation records, and preserve human oversight at every consequential step. As it turns out, those same properties are also what separate AI recruitment tools that make hiring better from the ones that just make it faster.
Ready to transform your hiring strategy? Schedule a Demo with our founders today!
Badis Zormati
Co-Founder, Asendia AI

