As artificial intelligence becomes more deeply embedded in everyday life, one question keeps coming back: how do we make sure AI systems are not only legal, but also fair, inclusive, and trustworthy?
With the release of Deliverable D3.2 – Ethical Considerations for AI Compliance, the CERTAIN Project takes an important step toward answering that question. This deliverable moves beyond high-level principles and shows how AI ethics can be translated into concrete, measurable, and auditable practices.
From principles to practice
Ethical AI is often discussed in abstract terms: fairness, transparency, accountability, human oversight. While these values are widely accepted, they can be difficult to apply in real-world AI development.
Deliverable D3.2 bridges this gap. Building on CERTAIN’s legal analysis (D3.1), it introduces a practical ethical governance framework that integrates ethics and gender considerations throughout the entire AI lifecycle, from design and development to deployment, monitoring, and post-market oversight.
Rather than treating ethics as an add-on, CERTAIN embeds it directly into everyday project workflows.
Four tools that make ethics actionable
At the heart of D3.2 are four practical instruments designed to help AI developers and organisations identify risks early and demonstrate responsible behaviour:
- Ethical Self-Assessment Checklist
A structured tool that helps teams evaluate ethical maturity using a clear scoring system and risk matrix. It supports transparency, accountability, and continuous improvement. - Gender Audit Tool
A method to assess whether gender equality and inclusiveness are truly embedded in data, design choices, and organisational practices, not just mentioned in policy documents. - Ethical and Gender Risk Register
A living register that records, tracks, and monitors ethical and gender-related risks, helping teams document mitigation measures and responsibilities in an auditable way. - Sex and Gender Impact Assessment (SGIA)
A framework that systematically examines how AI systems may affect different genders, supporting inclusive design and preventing unintended bias or exclusion.
Together, these tools transform ethical reflection into structured decision-making, making it easier to detect risks, prioritise actions, and show compliance.
Ethics, gender, and the EU AI Act
With the EU AI Act setting new expectations for trustworthy AI, ethical readiness is becoming a key part of compliance and future certification. Deliverable D3.2 aligns closely with the Act’s requirements on human oversight, transparency, fairness, and accountability.
By integrating ethical and gender governance into risk management and documentation processes, CERTAIN helps ensure that AI systems are not only technically robust, but also socially responsible and regulation-ready.
Importantly, this approach recognises that trust in AI depends not just on performance, but on how systems affect people, especially those who are often underrepresented or disproportionately impacted by automated decision-making.
Building trust by design
D3.2 reinforces CERTAIN’s broader vision: making human-centric AI the default, not the exception. By operationalising ethics and gender considerations, the project demonstrates how values can be embedded into real systems, rather than remaining theoretical ideals.
As AI regulation continues to evolve, CERTAIN’s ethical governance framework offers a practical roadmap for organisations that want to innovate responsibly, earn public trust, and prepare for future certification schemes.