The European standards we speak

We publish machine-readable model cards for the matching and placement models, store a factor-by-factor explanation for the top results of a talent-match search, and expose it to the student at /matches/[id]/why. Sensitive operations are written to an audit log.

The algorithm registry and per-match explanations are live on the talent-match search. Bias testing, coverage of the other scoring systems and Annex IV technical documentation do not exist.

Self-service exposure of Art. 15/16/17/20 rights. Per-category cookie consent. Primary database in the European Union; the privacy page lists which sub-processors sit outside the EEA and on what transfer basis.

Access, rectification, erasure and portability are self-service. Art. 22 human review is requested in-product from the match explanation: the request is recorded with a reference and a 30-day due date, and when it is closed the reviewer's outcome is written back onto the explanation. Signed-out visitors and the algorithm registry still fall back to email.

Today: an audit log on sensitive operations, a public model-card registry, and a required human decision in Talent Suite screening. Bias testing and data-governance documentation do not exist.

We are not certified to ISO/IEC 42001 and we do not have a documented AI management system.

Skill terms resolve to ESCO URIs through our SkillMapping table first, then a small curated map, then the live ESCO API, with hits cached back into SkillMapping. Matching unifies skill terms through that ESCO-mapped index. The match score itself carries no ESCO URI.

Students export their evidence-backed profile as Europass v3 JSON-LD with a single click.

Every Verifiable Credential we issue is exportable in EDCI JSON-LD format with the Ed25519 proof preserved. The single-credential export also carries the EQF level held on the student's record, when there is one.

Credentials are issued in the W3C Verifiable Credentials document shape and signed with Ed25519 over a canonical JSON serialisation of our own — not the W3C Ed25519Signature2020 suite, which requires URDNA2015 RDF canonicalisation and a multibase proofValue. The public key is published for offline verification.

Students export their Advanced/Expert skills as Open Badges 3.0 credentials.

Digital competences in our catalogue use DigComp's names and areas. The identifiers we assign are not URIs published by the JRC.

Entrepreneurship competences use EntreComp's names and areas, with the same limits on identifiers.

Sustainability competences use GreenComp's names and areas, with the same limits on identifiers.

EQF levels are assigned by the awarding authority, not derived by us. The single-credential EDCI export carries only a level already on the student's record, and omits the field when there is none.

A placeholder SAML metadata endpoint exists and returns an empty descriptor. The authentication flow is not built.

SPID/CIE scaffold: typed IdP list and attribute model. No login flow is wired; AgID registration is pending.

The EDCI and W3C VC credentials we issue are in the formats EUDIW wallets are expected to read; no wallet integration has been built or tested yet.

Credentials are EDCI-ready today. Wallet integration ships when reference implementations reach public beta.

The data model for ECTS equivalences exists, but no screen uses it today.

Exporting exchange enrolments in OLA JSON-LD exists as an API endpoint; no screen calls it today. EWP network integration requires per-institution SAML onboarding.

The data model for bilateral agreements exists; no screen uses it today.

Not connected to the EWP network; no endpoint is registered.

There is no integration with the ENIC-NARIC network.

Planned, with no date.