EU AI Act Article 86: The Right to Explanation Your HR Team Must Know

# The Rule Is Simple: Affected People Can Ask Why

The AI Act Article 86 right to explanation gives any person who is subject to a decision made with a high-risk AI system the right to ask the deployer for a clear, meaningful explanation of how that decision was reached. This is not a vague aspiration buried in a recital. It is a binding obligation that applies to you as a deployer the moment you use qualifying AI in contexts like recruitment, employee evaluation, credit scoring, or benefits assessment.

The rule came into force on 2 August 2026, the date by which most high-risk deployer obligations under Chapter III, Section 4 of the AI Act become applicable. If your company uses an AI-assisted applicant tracking system, an automated CV screener, or a performance scoring tool, you are already in scope.

# Who Is Covered and When

Article 86 applies to natural persons. That means employees, job applicants, customers, and any individual whose situation is materially affected by an output from a high-risk AI system listed in Annex III of the AI Act.

The scope is broader than many compliance officers initially expect. Annex III covers:

• Recruitment and selection of natural persons (including CV screening, ranking, and interview analysis tools)
• Decisions affecting terms of employment, promotion, and termination
• Access to education and vocational training
• Creditworthiness assessment of natural persons
• Benefits and social services eligibility decisions

A decision does not need to be fully automated to trigger the right. If the AI output materially influences a human decision-maker, the right to explanation still applies. That is explicit in Recital 86 of the AI Act, which clarifies that the explanation covers the role the AI played in the overall decision-making process.

Timing matters too. The right arises after a decision is communicated. There is no pre-decision consultation right under Article 86, though GDPR Article 22 may impose separate requirements for fully automated decisions. The affected person can exercise their right within a reasonable period following notification.

# What the Explanation Must Actually Contain

The AI Act does not let you satisfy this obligation with a boilerplate letter. The explanation must be meaningful and specific to the individual. Based on Article 86 and the accompanying recitals, it must address:

1. The main factors that influenced the AI system's output
2. The relative weight of those factors in the specific case
3. The logic the system applied, in plain language
4. How the AI output fed into the final decision

You do not need to disclose trade secrets or proprietary model weights. But you cannot hide behind vendor opacity. As a deployer, you must obtain sufficient information from your provider under Article 13 (transparency obligations) and Article 26 (deployer duties) to construct a genuine explanation. If your vendor cannot provide that, that is a procurement problem you need to solve now.

# Article 86 and GDPR: They Work Together

If you operate in the Netherlands or Belgium, you already know the Autoriteit Persoonsgegevens and the APD/GBA take automated decision-making seriously under GDPR Article 22. Article 86 of the AI Act runs alongside that framework, not instead of it.

GDPR Article 22 covers fully automated decisions with legal or similarly significant effects. The AI Act's Article 86 right applies more broadly, including decisions where a human is nominally involved but the AI output is the primary driver. In practice, for HR teams, this means:

• A rejected job applicant may have both a GDPR Article 22 right (if the rejection was fully automated) and an AI Act Article 86 right
• The explanation you provide must satisfy both regimes
• Your data protection officer should be involved in designing the template responses

Keep documentation. Both the GDPR and the AI Act require you to demonstrate compliance. A contemporaneous record of each explanation provided is your evidence in the event of a complaint to a supervisory authority.

# What a Non-Compliant Response Looks Like

Here is the kind of response that will create problems:

"Your application was reviewed using our standard recruitment process. We received many strong applications and have decided to proceed with other candidates."

This tells the applicant nothing about the AI's role. It does not acknowledge the system was used. It does not explain which factors mattered. It is not compliant.

Penalties under Article 99 of the AI Act for violations of deployer obligations can reach €15 million or 3% of global annual turnover, whichever is higher. For a 50-person company with €5 million turnover, that is a €150,000 exposure. Not a theoretical risk.

# Sample Response Templates for HR Teams

You should have at least two templates ready: one for rejected applicants and one for employees subject to performance or promotion decisions.

Template 1 — Rejected Applicant

Dear [Name],

Thank you for applying for the role of [Position]. We want to be transparent about our process.

We used an AI-assisted screening tool during initial shortlisting. In your case, the system assessed the following factors: [list, e.g. relevant experience duration, keyword alignment with the job description, qualifications match]. The factor that carried the most weight in your profile was [specific factor]. The system's output indicated [outcome, e.g. a below-threshold match score] and this output informed our recruiter's decision not to advance your application.

If you would like to discuss this further or provide additional context, please contact [name/email] within 30 days.

Template 2 — Employee Performance Decision

Dear [Name],

As discussed on [date], your [performance review / promotion assessment] outcome was [result]. We are writing to confirm the role our AI-assisted evaluation tool played in that process.

The tool assessed [factors, e.g. output metrics over the review period, attendance data, peer review scores]. The primary factors influencing the AI's output were [list]. Our [manager title] reviewed this output and made the final decision, taking [additional context] into account.

You have the right to request further information about this process. Please contact [HR contact] if you wish to do so.

Adapt these to your specific system and sector. The key requirement is specificity: name the factors, name the weight, name the system's output.

# Three Practical Steps to Get Ready

Step 1: Map your AI systems against Annex III. Identify every tool your company uses in HR, credit, benefits, or access decisions. If it scores, ranks, or recommends decisions about people, it likely qualifies.

Step 2: Talk to your vendors. Request documentation of the factors and weights their systems use. Ask them explicitly: can you support our Article 86 obligations? Get the answer in writing and make it a contractual term going forward.

Step 3: Prepare and test your templates. Draft responses for each AI-assisted decision type, run them past your DPO or legal counsel, and train the HR staff who will send them. Do a dry run with a fictional applicant to find gaps before a real request arrives.

Article 86 is one of the clearest, most enforceable individual rights in the AI Act. It is also one of the most actionable for HR teams. The window to prepare is open now.

Run the free 2-minute compliance check at comply.khairos.ai to see which Article 86 obligations apply to your specific AI tools and get a prioritised action list tailored to your company size and sector.