The standards behind the seat

We are a talent business first. We place, share, and grow AI ethicists; "evidenced against what" is the first honest question any talent organisation should ask, so here is the actual landscape a working AI ethicist draws on, not just the one spec that is easiest to put on a job ad.

The policy layer

The policy layer is two documents that set the tone globally without being law anywhere on their own: the UNESCO Recommendation on the Ethics of Artificial Intelligence and the OECD AI Principles. Between them they set out most of the substance everything else borrows from.

• The OECD frames it as human-centred design: systems built around dignity, individual autonomy, non-discrimination, and privacy, not just around what the technology can do, alongside transparency and explainability, robustness and safety, and accountability.
• UNESCO adds a wider duty to human well-being and the environment, and both are explicit that the aim is not only to stop AI doing harm, it is for AI to contribute something good, and for people to keep the right to decide who gets to decide.

Neither document is enforceable in the way a regulation is, but both shape how boards, funders, and regulators talk about AI, and a practitioner who only knows the two names, not what is actually inside them, is going to sound out of step in most serious rooms.

The regulatory layer

The regulatory layer is for example the EU AI Act, the one document here with actual legal teeth for organisations in scope: obligations, risk tiers, and deadlines that do not move because a policy paper was not binding. An ethicist does not need to practise law, but working fluency here, knowing what applies, when, and to what, is table stakes for the role, not a nice-to-have.

The competence layer: EN 18274

EN 18274 is the closest thing this profession has to a shared job description, published by CEN in 2026.

What follows is our own reading of what EN 18274 sets out, built for hiring conversations, not a substitute for the published text, which any organisation relying on it formally should read directly. As we would describe it: the knowledge side is substantive, not procedural, recognising fairness and bias in a live system, explainability, privacy, human autonomy and oversight, safety and robustness, and where accountability actually sits, not just the names of these ideas.

The practice side is how that knowledge gets used day to day, roughly: governance-facing work, stakeholder engagement, telling ethical concerns apart from legal ones, communication, and system-facing practice, embedding ethics into how AI actually gets built, running impact assessments, documenting reasoning, mediating disagreement. That depth is what makes it useful in a hiring conversation, even though it is not, yet, a certification anyone can sit and pass; no accredited scheme exists against it anywhere.

What if someone already holds a certification?

Fair question, since EN 18274 has no certification scheme yet and IAPP's AIGP (Artificial Intelligence Governance Professional) does. It is a real, sittable credential, and a meaningful share of the practitioner pool already holds it. We treat it as strong evidence, not a substitute for the conversation: AIGP tells you someone has passed a broad governance exam, it does not tell you how they handled a launch date nobody wanted to move. We ask both. Sentient Staff is not affiliated with or endorsed by IAPP.

The organisational layer: ISO/IEC 42001

Everything above is about the person. ISO/IEC 42001 is about the organisation, a management-system standard for how a company actually runs its AI governance, closer in spirit to ISO 27001 for information security than to a competence framework for individuals. A company can be certified against it; a person can't. We mention it because a hiring organisation working toward ISO/IEC 42001 needs an ethicist who understands what that certification actually requires operationally, not someone who's only heard the acronym. Sentient Staff is not affiliated with or endorsed by ISO.

The standards most people have never heard of

Beyond all of the above, a handful of quieter standards shape how a serious engagement actually gets run: guidance on running a proper AI impact assessment, guidance on building a whistleblowing system that actually protects the person who raises a finding, and a broader European ethics framework written for the ICT profession generally rather than AI specifically. None of these get quoted in job ads. All of them show up in the work.

Why we do not peg the business to just one of them

We do not peg every placement to a single named standard because no one standard covers the job. The engagements we run evidence a person's competence against whichever combination of these actually applies to the role in front of us, sector, geography, and risk profile all change the mix, with the reasoning shown rather than a badge asserted. That is a professional judgement, not a certificate, and it is a more honest description of what this work actually is than any one acronym on its own.

Where the profession stands, more broadly

Ethics Directive, a separate register tracking AI ethics credentials and standards, is building an annual, aggregated report on where the AI ethicist role actually stands across sectors. It is is worth watching: the state of the AI ethicist.