EU AI Act Compliance

Building AI that
earns trust.

pace is classified as a high-risk AI system under the EU AI Act. We embrace this classification — it reflects the gravity of the decisions our technology supports.

Classification

High-risk AI —
by design, not by accident.

Under Article 6 and Annex III of the EU AI Act, AI systems used in employment, worker management, and access to self-employment — including recruitment and selection — are classified as high-risk. pace falls squarely within this category, and we've built our architecture to meet that standard from the ground up.

01

Human-in-the-Loop Design

pace never makes hiring decisions. Every verdict — hire, advance, or decline — is made by a human interviewer. The AI provides structured evidence, suggested questions, and calibrated scoring, but the final judgment always belongs to the person conducting the interview.

02

Bias-Protected Score Isolation

Individual interviewer scores are isolated until the full panel has submitted their evaluations. This prevents anchoring bias, groupthink, and social conformity from contaminating the signal. Each assessment stands on its own merit before aggregation reveals the collective view.

Transparency Measures

Explainable, auditable,
human-centred.

01

Human-in-the-Loop Design

Every verdict—hire, advance, or decline—is made by a human interviewer. The AI provides guidance throughout the process by offering structured evidence from the interview, high-quality personalized suggested questions, and calibrated scoring, among other features. The final judgment always belongs to the person conducting the interview.

02

Bias-Protected Score Isolation

Individual interviewer scores are isolated until the full panel has submitted their evaluations. This prevents anchoring bias, groupthink, and social conformity from contaminating the signal. Each assessment stands on its own merit before aggregation reveals the collective view.

03

Explainable AI Signals

Every score, suggestion, and competency evaluation generated by pace can be traced back to specific behavioral evidence observed during the interview. There are no black-box outputs — every signal has a rationale, and every rationale references observable data.

04

Structured Competency Frameworks

Scoring is anchored to pre-defined competency frameworks with behavioral indicators at each level. This ensures consistency across interviews, reduces subjectivity, and creates an auditable record of what was evaluated and why.

Data Governance

Governance isn't overhead.
It's infrastructure.

01

Data Quality

Structured inputs, verified outputs.

All data entering the pace system passes through validation layers. Interview evidence is structured, timestamped, and linked to specific competency dimensions. We maintain data lineage so every output can be traced to its source.

02

Monitoring

Performance oversight in development.

We are building capability to monitor model outputs for drift, bias patterns, and anomalies. Automated bias drift monitoring and regular auditing of scoring distributions across demographic groups are on our roadmap — this is planned, not yet operational.

03

Documentation

Working toward Article 11 compliance.

The EU AI Act requires comprehensive technical documentation covering system architecture, performance metrics, known limitations, and intended use cases. We are developing this documentation and plan to make it available to competent authorities as required. This is a work in progress.

Conformity Approach

Our path to full
regulatory alignment.

An honest view of where each milestone stands on the path to full EU AI Act conformity.

1

Risk Assessment

In Progress

Systematic identification and evaluation of risks to fundamental rights, including potential for discrimination, privacy impacts, and effects on employment decisions. This is an ongoing effort as we refine our risk management approach.

2

Quality Management

Developing

We are developing documented quality management practices covering development, testing, and validation. Formalised quality management systems are on our roadmap as we prepare for future certification.

3

Conformity Assessment

Planned

Internal conformity assessment procedures aligned with Annex VI of the EU AI Act are planned. We have not yet completed a formal conformity assessment or engaged third-party auditors for this purpose.

4

EU Registration

Planned

Registration in the EU database for high-risk AI systems as required by Article 49 is a future milestone. We will register when the regulatory framework and database are fully operational and applicable to our deployment.

Frequently Asked

What HR and compliance teams ask us about the EU AI Act.

Is pace classified as a high-risk AI system under the EU AI Act?

Yes. We self-classify under Article 6 and Annex III, point 4 of the EU AI Act, which covers AI systems used in employment, worker management, and access to self-employment — explicitly including recruitment and selection. Rather than try to argue our way out of that classification, we have built the platform to meet the obligations it brings: human-in-the-loop decisions, traceable scoring, structured competency frameworks, and a roadmap toward formal conformity assessment.

What does Article 26 mean for us as the deployer using pace?

Article 26 places obligations on you, the deployer, when you use a high-risk AI system in HR. In short: you must use the system in line with its intended purpose, assign human oversight to people with the competence to provide it, monitor its operation, keep logs for an appropriate period, and inform candidates that an AI system is being used in the selection process. pace supports this in concrete ways: scoring and recommendations are presented to a named human reviewer rather than auto-applied, every score links back to the evidence that produced it, the platform retains a structured record of who reviewed what and when, and — because an automated notetaker joins the interview to capture the transcript — we provide notice copy you can include in candidate-facing communications so that you can inform candidates of the recording and AI use before the interview. We will share our internal documentation under NDA to help you build your own Article 26 file.

How do you test for bias in your scoring models?

Three layers, today. First, scoring is anchored to behavioural indicators in your competency framework rather than to free-form impressions, which removes a large class of subjective drift before it starts. Second, every AI-generated score runs through a reflexion pass — a second model critique that checks the score against the supporting evidence and flags weak rationales for human review before they reach the panel. Third, edits to competency frameworks and scorecards are written to our internal audit log, so changes that could shift outcomes for protected groups are traceable. We are honest that systematic, statistical bias-drift monitoring across demographic groups is on our roadmap and not yet operational at scale; we are building toward it as our customer base grows large enough to make those statistics meaningful.

Next Steps

Ready to see compliance
in action?

Experience how pace builds trust into every hiring decision — with human oversight, bias protection, and explainable AI at every step.