AI systems used in recruitment and employment are explicitly classified as high-risk under the EU AI Act. Most Irish employers using AI in hiring have not assessed their obligations. Stanford research shows screening tools discriminate. Here is what needs to change.
The EU AI Act is unambiguous on this point: AI systems used in recruitment, hiring, and employment decisions are classified as high-risk. This classification carries the Act's most demanding compliance requirements — conformity assessments, human oversight mechanisms, transparency obligations, bias testing, and documentation.
Most Irish employers using AI in their recruitment processes have not assessed their obligations under this classification. Many do not realise they are deploying high-risk AI systems at all.
What counts as high-risk HR AI
The scope is broader than most employers expect. High-risk classification applies to AI used for:
- CV screening and candidate ranking — automated systems that filter or score applications
- Video interview analysis — tools that assess candidate suitability from recorded interviews
- Automated shortlisting — any system that determines which candidates progress
- Performance evaluation — AI-assisted performance scoring or review systems
- Promotion decisions — tools that identify or rank employees for advancement
- Workforce monitoring — AI systems that monitor employee behaviour or productivity
If your organisation uses any of these — including through third-party recruitment platforms or ATS systems with embedded AI features — you are deploying high-risk AI under the EU AI Act.
The bias evidence is clear
Stanford researchers published findings in October 2025 showing that AI resume-screening tools gave older male candidates higher ratings than both female candidates and younger candidates — despite all candidates' resumes being generated from identical data sets. The bias was not in the data the researchers provided. It was in the AI model's training data and evaluation patterns.
This is not an isolated finding. AI-powered hiring tools processed over 30 million applications in 2024 alone, triggering hundreds of discrimination complaints. The pattern is consistent: AI recruitment tools can reproduce and amplify human bias. If the training data reflects historical hiring patterns that favoured certain demographics, the AI tool will replicate those patterns — with the added authority of appearing objective.
For Irish employers, this has specific implications. The Employment Equality Acts already prohibit discrimination in recruitment. The EU Pay Transparency Directive adds requirements around objective, gender-neutral criteria in hiring and pay decisions. AI tools that introduce or amplify bias expose the employer to legal challenge under both frameworks.
What the EU AI Act requires for HR AI
As a deployer of high-risk AI in employment, Irish organisations must:
Ensure human oversight. Every AI-assisted hiring decision must have meaningful human review. "Meaningful" means a trained person who is empowered to override the AI — not a rubber stamp.
Conduct bias testing. AI recruitment tools must be tested for discriminatory outcomes across protected characteristics. This is not a one-off exercise — it requires ongoing monitoring.
Maintain transparency. Candidates must be informed when AI is used in the hiring process. Under the Pay Transparency Directive, salary history requests are prohibited and pay ranges must be disclosed — removing two of the mechanisms through which AI-amplified bias typically manifests.
Document everything. The risk assessment, bias testing results, human oversight mechanisms, and governance structures must be documented and available for regulatory inspection.
Test vendor claims. If you use a third-party recruitment platform with AI features, the vendor's assurances about bias and fairness are not sufficient. You — as the deployer — carry the compliance obligation. Vendor due diligence must be documented and rigorous.
The intersection with pay transparency
There is a direct connection between AI in recruitment and pay transparency compliance. If AI tools influence who gets hired and at what salary level, and those tools contain bias, the resulting pay structures will embed that bias structurally.
The Pay Transparency Directive requires employers to demonstrate that pay differences between comparable roles are justified by objective, gender-neutral criteria. If AI screening tools systematically place women or older candidates in lower salary bands — as the Stanford research suggests — the employer's pay equity position is compromised before the employee has even started.
This is why job architecture and AI-assisted role analysis matter. They provide the objective framework against which pay decisions can be assessed — independent of the biases that may exist in recruitment AI.
What Irish employers should do now
1. Inventory your recruitment AI. Every AI tool used in hiring — including features embedded in your ATS, video interview platforms, and assessment tools — needs to be identified and classified.
2. Assess vendor compliance. For third-party tools, conduct due diligence on bias testing, data practices, and the vendor's own EU AI Act compliance position. If the vendor cannot provide adequate documentation, that is a red flag.
3. Establish human oversight. Ensure that every AI-assisted hiring decision has genuine human review by someone with the authority and training to override the AI's recommendation.
4. Connect to your pay transparency work. If you are building a job architecture for pay transparency, ensure that the recruitment process feeds into it — with salary placement based on objective criteria, not AI-influenced starting points.
5. Brief your board. HR AI risk should be on the board's agenda. The combination of employment law, the EU AI Act, and the Pay Transparency Directive creates a compound regulatory exposure that requires board-level oversight.
If your organisation uses AI in recruitment and has not assessed its obligations, contact Acuity AI Advisory for a structured assessment covering EU AI Act classification, bias risk, and pay transparency intersection.