Your AI Resume Screener Is Already in the EU's Sights: What to Change | inicIA

Direct answer

The EU AI Act classifies AI systems for recruitment, screening, and candidate evaluation as high-risk: fines for non-compliance reach up to EUR 15 million or 3% of global annual turnover. Two dates you must know: Article 50 is already in force since August 2026 (if a candidate interacts with an AI interviewer in the EU, you must tell them they are talking to an AI), and full high-risk obligations arrive on December 2, 2027 (the Digital Omnibus deferred them from August 2026). On top of that, New York's law (Local Law 144, in force since 2023) already requires independent bias audits for employment decision tools. If you use AI to screen resumes or evaluate candidates, 2026 is the year to fix that process — not to wait for 2027.

Why hiring is the most watched front

Regulators target recruitment first because the harm is measurable and the track record is bad. Two cases you should know:

Add Colorado (SB 26-189, in force since January 2027): advance notice, adverse-decision explanations within 30 days, and human reconsideration. The pattern is global: hiring with opaque AI is becoming legally expensive.

What already binds you today (Article 50)

Since August 2026, the transparency duty is in force and was not deferred: any candidate in the EU interacting with an AI interviewer or evaluator must be told they are dealing with an AI, and generated content must carry machine-readable marking. This applies based on where the candidate is, not where your company sits. If your video-interview tool serves European candidates, this already reaches you.

The 5-step plan before December 2027

  1. Inventory. List every AI tool touching your hiring process: resume filters, video interviews with analysis, automated tests, scoring. Include the ones embedded in your ATS.
  2. Demand the paperwork from vendors. For each high-risk tool, require: risk assessment, technical documentation, human-oversight plan, and instructions for use. Without vendor paperwork, you carry the risk.
  3. Real human oversight. The law requires a person with authority to intervene in, review, and override the AI's decision. "The AI decides and the human signs" does not comply.
  4. Bias testing. Run a bias audit on your tools (as New York requires): compare outcomes across demographic groups using synthetic or anonymized historical data.
  5. Candidate notice. Document what the candidate is told, when, and how: what the AI decides, what data it uses, and how to request human review. Article 50 already requires this.

Related articles

Frequently asked questions

Do hiring fines already apply?

Transparency (Article 50) does, since August 2026. Full high-risk obligations (risk management, documentation, conformity assessment) arrive December 2, 2027 after the Digital Omnibus deferral. Fines up to EUR 15M or 3% apply to that block.

I only hire in Miami. Does it affect me?

If no candidate is in the EU, the AI Act does not reach you through that channel. But New York and Colorado already legislate, and your large clients will ask for assurances. Getting in order now is a commercial advantage.

What if my vendor says its tool "has no bias"?

Ask for it in writing, with its audit methodology and results. A verbal claim is not documentation, and the law requires documentation.

Is emotion recognition in interviews allowed?

No: the AI Act classifies it as a prohibited practice in work and education contexts, with fines up to EUR 35M or 7%.

Take the next step

If you use AI in hiring — or plan to — and want a process that complies without slowing your recruiting, talk to inicIA on WhatsApp: https://wa.me/17866120930

By inicIA