About Ethics Directive

The assay is the test. The hallmark is only the mark that follows it.

For six hundred years, before you trusted a piece of silver, someone scraped it, dissolved a little in acid, and weighed what was left.

That was the assay. Only then did the object get its mark. The mark was never the point. The mark was shorthand for a test that had already happened, performed by someone with no stake in the sale.

Europe is about to start marking people. EN 18274 sets out what a competent AI ethicist knows and can do, and in time there will be accredited bodies handing out certificates against it. That is a good thing, and overdue. It also raises the oldest question in the assay office: the mark tells you a test happened, so who did the test, and can you trust them?

That is what we are here for. We called ourselves Ethics Directive because that is the register this field already speaks in: a directive is the thing you are measured against, and a hallmark is what says you met it. We are the independent read on both, the standard behind the mark, and whether the mark is worth anything.

What we do

A note on names first. Some people call this field AI ethics; others call it ethics for AI or ethics of AI; regulators mostly say trustworthy AI. Same discipline, same register, and we answer to all of them.

We tell you, plainly and for free, what an AI credential is actually worth. There are already more than a dozen of them, each from a different tradition, each with its own price, its own entry requirements, and its own quiet assumptions about what “ethics” means. There was no map, so we built one. We keep four things, starting with the one we are named after:

• A register: what EN 18274 is, where it stands, and the honest answer to “can anyone certify me yet?”
• A map: every credential in the neighbourhood, what it costs, what it gates, and what it is worth.
• A statute: the wider law and policy an AI ethicist answers to beyond the standard itself, such as the EU AI Act, NIST’s AI RMF, the OECD AI Principles, UNESCO’s Recommendation on the Ethics of AI, and the rest, binding or not, mapped in one place.
• The minutes: a tool that reads what you type and tells you which excuse the room used, built on real accounts of the day someone said no.

Who this is for

The person deciding whether a course is worth their evening, weekend, or holiday. The head of legal who wants a straight answer for the board on what the AI ethicist they just hired is actually qualified to do. The board that appointed an ethicist and wants to know the role has been set up to carry real weight. And the practitioner who raised an objection that mattered, and wants to know their judgement was sound, whatever the room decided at the time. If you saw yourself in any of those four, good, you were meant to.

Who runs this

Currently: a small team of AI ethicists, which, per our own register, nobody can technically be certified as yet. We are working on that too.

What we don’t know

Quite a lot, and we would rather say so. At the time of writing, we do not know which accreditation body will move first, or when the first genuine AI Ethicist certificate will exist. Where the honest answer is “nobody, yet”, that is what the register says. Being specifically ignorant in public tends to pull the truth out of people, which is rather the idea.

How we pay for this

Here is the part most people bury, so we lead with it. Practitioners, employers, and public grants fund the work; the credential owners we assess do not.

We know how the other version ends: a rater paid by the people it rates tends to find good things to say about them, however careful its conflicts policy. Disclosure is not a solvent, so we built the structure differently instead. The full breakdown lives here.