Available in 2026
CEN lists EN 18274:2026 as published. National publications or adoptions may follow the European date of availability.
AI governance has spent years defining principles, controls and system requirements. EN 18274:2026 adds a different question: who is competent to help an organisation identify, deliberate on and manage the ethical and societal implications of AI?
The standard creates a common European reference for the professional AI ethicist role. Its published scope organises competence around knowledge, skills and attitudes and addresses how the role can be integrated internally or externally in commercial, government and non-profit organisations.
That is more useful than inventing an impressive job title. It gives organisations a basis for designing responsibilities, selecting internal or external expertise, planning professional development and asking for evidence that competence can be demonstrated in practice.
Publication status matters
CEN’s project record lists EN 18274:2026 as published, with a date of availability of 5 August 2026. It also lists 30 November 2026 for national announcement and 28 February 2027 for national publication and withdrawal of conflicting standards. Some catalogues still display the earlier prEN 18274:2025 draft. For implementation or assurance work, use the definitive EN text or its official national adoption—not an earlier draft.
From ethical principles to demonstrated competence
Many organisations have AI principles. Fewer can show who turns those principles into decisions when a real system creates tensions between accuracy, privacy, fairness, safety, sustainability, commercial value and human autonomy.
EN 18274 moves the discussion from abstract values toward professional capability. It treats competence as demonstrated performance, not knowledge alone.
The final CEN scope confirms that high-level framework. The more detailed capability map below summarises the earlier prEN 18274:2025 development text. It is useful for orientation, but it is not a substitute for checking the definitive EN text.
In practice, that means an AI ethics role should be able to contribute to work such as:
- explaining ethical and societal issues to technical and non-technical audiences;
- engaging affected stakeholders and surfacing conflicting interests;
- distinguishing legal requirements from wider ethical concerns;
- integrating ethical considerations into AI design, deployment, monitoring and retirement;
- supporting ethical and societal impact assessments;
- examining whether governance practices and AI systems align with stated commitments;
- documenting deliberation, trade-offs, residual concerns and decisions;
- facilitating difficult conversations without pretending that one expert owns the final answer.
Two practical layers from the development text
The earlier prEN development text grouped the work into two connected layers. They remain a useful lens for organisational preparation, while detailed implementation should be checked against the definitive standard.
1. Governance competence
This layer concerns the organisation around the technology: awareness and education, stakeholder management, communication, the boundary between ethics and law, and the resources required to maintain an effective ethics function.
2. AI system life-cycle competence
This layer brings ethics into the work itself: implementation, impact assessment, ethics-based audit, documentation quality, ethical reasoning and mediation across design, development, deployment, monitoring and retirement.
A role that only discusses values but cannot enter product, procurement or risk workflows is too detached. A role that only completes compliance checklists may miss legitimate concerns that sit beyond minimum legal requirements. Credible AI ethics needs both layers.
The following table is Ada Studio’s implementation aid. It is not a requirements table from EN 18274.
Training
Organisational need: Role-based awareness and education
Evidence of useful competence: Materials adapted to audience, workflow and risk; learning outcomes and refresh triggers
Stakeholders
Organisational need: Structured engagement
Evidence of useful competence: Stakeholder map, documented concerns, feedback channels and escalation routes
Ethics and law
Organisational need: A clear boundary between the two
Evidence of useful competence: An account of what is legally required, ethically contested and still undecided
AI life cycle
Organisational need: Ethics integrated into delivery
Evidence of useful competence: Defined review points in design, procurement, deployment, monitoring and retirement
Assessment
Organisational need: Impact assessment and ethics-based audit
Evidence of useful competence: Traceable method, evidence, findings, limitations and improvement actions
Deliberation
Organisational need: Structured ethical reasoning
Evidence of useful competence: Options, affected values, trade-offs, residual concerns, decision owner and rationale
The AI ethicist is an adviser—not the organisation’s moral judge
One of the most important governance principles is role clarity. Ethical decisions cannot be outsourced to a single specialist. The AI ethicist should identify concerns, improve the quality of deliberation, challenge weak assumptions, facilitate stakeholder dialogue and support defensible decisions. Accountability remains with the relevant leaders and decision-makers.
An AI ethicist should improve the quality of decisions—not absorb accountability for them.
This has design consequences. The role needs enough independence to raise uncomfortable issues, confidentiality to handle sensitive information, access to relevant technical and business evidence, and an escalation path when concerns are ignored. It also needs boundaries: ethics expertise does not replace legal, security, safety, data protection, engineering or domain expertise.
Four ways to organise the capability
CEN’s published scope allows the role to be fulfilled internally or externally. The standard should therefore not be read as requiring every organisation to create a full-time post. The right model depends on AI maturity, risk profile, organisational size and available resources.
The options below are Ada Studio’s practical role-design patterns, not operating models prescribed by EN 18274.
- Embedded specialist: an AI ethics professional works inside product, data, innovation, legal or risk teams. This creates proximity to decisions but requires safeguards against conflicts of interest.
- Central ethics or responsible-AI function: a dedicated team provides methods, challenge and oversight across multiple business units. This supports consistency but can become distant from delivery.
- External adviser or advisory board: independent expertise is used for defined assessments, high-stakes decisions or periodic review. This can work well for smaller organisations if access and continuity are designed properly.
- Hybrid model: internal owners manage daily governance while external specialists provide independent challenge, facilitation and specialist depth.
The title matters less than the decision rights, access, competence and evidence behind it.
Seven questions to ask before defining the role
AI ethics capability design checklist
0/7Ada Studio’s suggested 90-day response
This is an implementation sequence proposed by Ada Studio, not a timetable required by EN 18274.
Days 1–30: map decisions, not titles
Identify the AI use cases and decisions where ethical and societal questions arise. Map current owners, existing risk and compliance processes, affected groups, evidence gaps and escalation paths. Do not begin with an organisation chart.
Days 31–60: assess the competence gap
Compare the needed work with available internal capability. Separate general AI governance skills from specialist ethical reasoning, impact assessment, facilitation and audit competence. Decide what can be developed internally and where independent external support is appropriate.
Days 61–90: pilot the operating model
Choose one material use case. Define the AI ethics role, decision owner, review points, stakeholder inputs, documentation template and follow-up mechanism. Test whether the arrangement improves the decision—not only whether a meeting occurred.
What publication does—and does not—mean
EN 18274:2026 is a voluntary European standard. Its publication does not automatically create a legal duty to appoint an AI ethicist, certify an individual, or follow one organisational model. It also does not by itself establish presumption of conformity with the EU AI Act. At the time of writing, CEN’s project record lists no citation in the Official Journal of the European Union.
Do not turn a competence standard into a badge
A job title, short course or certificate is not enough on its own. Organisations should look for demonstrated ability, relevant experience, transparent limits, ongoing learning and evidence from real governance work.
The standard is nevertheless timely. For providers and deployers, the EU AI Act’s Article 4 AI literacy requirement, ISO/IEC 42001’s requirements for an AI management system and ISO/IEC 42005 guidance on AI system impact assessment all increase the need for people who can connect technology, governance, fundamental rights, organisational realities and stakeholder concerns. Article 4 does not itself require the appointment of an AI ethicist.
EN 18274 gives that capability a clearer professional shape. The opportunity now is to integrate it without creating a symbolic ethics role that sits outside real decisions.
Sources and status note
- CEN project record: EN 18274:2026 — authoritative final-project status; accessed 7 August 2026
- AI Standards Hub record for the earlier prEN 18274:2025 draft — third-party draft record; accessed 7 August 2026
- ISO/IEC 42001:2023 — AI management systems
- ISO/IEC 42005:2025 — AI system impact assessment
- Regulation (EU) 2024/1689 — consolidated EU AI Act text
Ada Studio helps organisations translate AI governance requirements into clear roles, practical controls, decision processes and role-based training. To discuss what EN 18274 could mean for your operating model, get in touch.