In this article9 sections
  1. Your ATS Is a High-Risk AI System. Full Stop.
  2. Provider Obligations: What Vendors Must Do
  3. Deployer Obligations: What Your Customers Must Do
  4. The Dual-Obligation Overlap
  5. Key Dates to Build Your Roadmap Around
  6. What to Tell Prospects Asking About Compliance
  7. The Discrimination Risk You Cannot Ignore
  8. One Action to Take This Week
  9. References

Your ATS Is a High-Risk AI System. Full Stop.

HR tech vendor EU AI Act compliance is not a distant consideration. The EU AI Act entered into force on 1 August 2024. For vendors selling AI-powered applicant tracking systems (ATS), candidate scoring engines, or performance management tools into the EU market, the obligations are real and tiered, and enterprise customers ask about them in procurement questionnaires.

The reason HR tech sits in the highest-scrutiny category is simple: Annex III of the EU AI Act explicitly lists AI systems used for recruitment and selection of natural persons, evaluation of candidates, and assessment of persons in employment relationships as high-risk AI systems. That classification does not depend on company size, revenue, or whether you are headquartered in Amsterdam or Atlanta. If your product is intended to filter applications, evaluate candidates, or support decisions about promotion, termination or task allocation, it is Annex III, point 4. Article 6(3) lets a provider conclude that an Annex III system is not high-risk only in narrow cases, and never when the system profiles natural persons, which is what scoring candidates usually involves.

Provider Obligations: What Vendors Must Do

As a vendor, the EU AI Act classifies you as a provider under Article 3(3). Provider status comes with the heavier burden. Since the Digital Omnibus on AI (Regulation (EU) 2026/1744), these obligations apply to Annex III systems from 2 December 2027. Here is the concrete list:

Conformity assessment. Before placing your high-risk AI system on the EU market, you must complete a conformity assessment. For HR tech providers, this is an internal-control procedure documented in a technical file, not a third-party audit, but that file must be comprehensive. It covers risk management, data governance, transparency, human oversight design, accuracy and robustness. SMEs and small mid-caps may provide the technical documentation in a simplified form (Article 11), and the quality management system may be proportionate to the size of the organisation (Article 17).

CE marking and EU declaration of conformity. Once conformity is assessed, you draw up an EU declaration of conformity (Article 47) and affix the CE marking (Article 48). From 2 December 2027, no CE marking means no lawful placing of a new high-risk HR system on the EU market.

Registration in the EU database. Providers must register high-risk Annex III systems in the EU database before placing them on the market. This is a public-facing step that your buyers will be able to verify.

Post-market monitoring. Article 72 requires providers to actively monitor system performance after deployment. You need a structured process to collect data on how your tool performs in real conditions, flag unexpected risks, and feed that back into the model.

Serious incident reporting. If your system is involved in a serious incident, for example an infringement of obligations under EU law that protect fundamental rights, such as a ban on discrimination, you must report it to the market surveillance authority under Article 73. Deployers who spot one must inform you first (Article 26(5)), so you need a channel for that.

Instructions for use. You must supply your deployers with clear instructions covering intended use, known limitations, required human oversight steps, and prohibited applications. Vague documentation is a liability, not a safe harbour.

Deployer Obligations: What Your Customers Must Do

Your EU customers, the HR teams and People Ops departments actually running your software, are deployers. Their obligations do not disappear just because they bought a compliant tool. Under Article 26, from 2 December 2027, they must:

  • Use the system in accordance with your instructions for use

  • Assign human oversight to people with the necessary competence, training and authority

  • Keep the logs the system generates for at least six months

  • Inform workers' representatives and affected workers before putting the system into use at the workplace, and inform candidates that a high-risk system is used in decisions about them

  • Inform you, and where needed the authorities, of risks and serious incidents

Separately, Article 4 has applied since 2 February 2025: they must take measures to support the AI literacy of the staff who use your tool. A private employer does not need a Fundamental Rights Impact Assessment; that duty is limited to public bodies, private providers of public services, and deployers of credit scoring or life and health insurance pricing AI.

Here is the commercial reality: deployers depend on your documentation to meet their own duties, and procurement teams increasingly include AI governance questions. Showing up to a sales conversation without answers means losing ground to vendors who have prepared.

The Dual-Obligation Overlap

Some vendors also deploy their own tools, for example if you run an AI recruiting platform as a service and actively match candidates to employer job posts yourself. In that scenario, you wear both hats simultaneously: provider obligations for the system, and deployer obligations for your own use of it. That means your internal compliance and your customer-facing compliance documentation are distinct deliverables.

The boundary gets blurry when you offer white-label or API-based products. If a customer puts its own name on your system, substantially modifies it, or builds a candidate-scoring workflow on top of a general-purpose component of yours, it can become the provider of that high-risk system under Article 25. Your contract language needs to specify the roles and the information you will provide. Ambiguity here creates legal exposure for both parties.

Key Dates to Build Your Roadmap Around

The EU AI Act applies in phases. The Digital Omnibus on AI moved the high-risk provisions for Annex III, the ones that directly affect HR tech, to 2 December 2027. That is a fixed date, not linked to the arrival of harmonised standards. It sounds like breathing room, but the conformity assessment, technical file preparation, and registration process are not weekend projects, and enterprise buyers tend to ask well before the legal date. Systems already on the market before 2 December 2027 only fall under the rules if they are significantly changed afterwards (Article 111(2)), but most SaaS products change constantly, so plan as if yours will.

The prohibited AI practices under Article 5 have applied since 2 February 2025. They include emotion recognition in the workplace (Article 5(1)(f)), except for medical or safety reasons. The Commission's guidelines read "workplace" as also covering candidates during selection and hiring, so an emotion-analysis feature in video interviews is prohibited: switch it off or remove it.

What to Tell Prospects Asking About Compliance

When an HR Director or Compliance Officer asks whether your tool is EU AI Act compliant, here is the framework for a credible answer:

  1. Confirm the classification. Tell them clearly that your system is classified as a high-risk AI system under Annex III, and that you are treating it accordingly. Downplaying the classification is a red flag to sophisticated buyers.

  2. Describe your conformity assessment status. If you have a completed technical file, say so. If you are in progress, give a completion date. Vagueness signals unpreparedness.

  3. Hand over the instructions for use. Proactively share the document that tells deployers how to implement human oversight. Buyers who receive this document without asking for it trust the vendor more.

  4. Clarify what they must do. Walk the prospect through their Article 26 deployer obligations. Explain that, as private employers, they do not need a FRIA unless they are a public body or provide public services. This positions you as a compliance partner, not just a software seller.

  5. Confirm your post-market monitoring process. Buyers want to know that if your model starts producing biased outputs in their environment, you will detect it and they will hear from you. Describe your monitoring cadence and escalation path.

This conversation, done well, shortens procurement cycles. Done poorly, it kills deals.

The Discrimination Risk You Cannot Ignore

HR AI sits at the intersection of the AI Act, the GDPR and equal-treatment law. Candidate scoring and performance assessment systems can reproduce discrimination by proxy, even if they never ingest protected attributes. Data protection authorities, including the Dutch Autoriteit Persoonsgegevens, already supervise the GDPR side of this.

For vendors, this means bias testing must be documented, not assumed. Article 10 requires providers of high-risk systems to examine training, validation and testing data for possible biases and take measures against them, and the new Article 4a allows the processing of special-category data for that purpose under strict conditions. Accuracy and fairness metrics by group, to the extent your data allows, are the practical evidence. If you do not have that documentation, build it before your next enterprise sales cycle.

One Action to Take This Week

Pull your product's current documentation and ask one question: could a deployer's Compliance Officer read this and know exactly how to implement human oversight, keep the logs and inform workers? If the answer is no, that document is your first priority. Rewrite it to meet the instructions-for-use standard in Article 13, so your customers can meet Article 26. That single deliverable will differentiate you in more sales conversations than any feature release this year.

Run your vendor compliance posture through the free 2-minute check at comply.khairos.ai to see exactly where your documentation gaps are before your next enterprise prospect asks.

References

  1. Annex III of the EU AI Act

  2. Article 3(3)

  3. Article 48

  4. Article 26

  5. Article 6

  6. Regulation (EU) 2026/1744 (Digital Omnibus on AI)

  7. comply.khairos.ai

Not sure where your company stands?

Answer 15 questions about how your team uses AI. You get a compliance score and the gaps to close first, as a PDF.

Am I compliant? Check in 2 minHow Khairos Comply works