AI Governance & Academic Oversight
How EANE uses AI to support human-led editorial work, where the boundaries are, and how the process is overseen. First formal external audit scheduled Q3 2027.
Human-led, AI-assisted, publicly accountable
EANE is a non-formal education provider building stackable European micro-credentials. AI plays a defined, supporting role in our editorial workflow: it accelerates translation, flags quality issues for human review, and drafts cover imagery. It does not, in any surface of our service, decide who has learned what.
This page sets out exactly what AI does at EANE, what it does not do, who is accountable, and when it is audited. If any of it is unclear or you want to raise a concern, email hello@eane.eu.
What AI does, and does not do, at EANE
What AI does
- •Translates editorially-approved English source content into EANE's other publishing languages, using a Google Gemini model.
- •Runs a Quality Audit pass over every published module, flagging ambiguous questions, weak distractors, tone drift and reading-level outliers for editorial review.
- •Suggests an indicative EQF (iEQF) level for each new module as a first-pass calibration, scored against the published rubric.
- •Generates cover imagery from editorial briefs, using a Google Gemini image model; every cover is selected by an editor before publication.
- •Reads text modules aloud via a text-to-speech pipeline, so learners can consume material on the move.
What AI does not do
- •Decide whether a learner has passed. Multiple-choice tests are graded against a fixed answer key set by the editorial team; no LLM is asked to judge learner responses.
- •Set final iEQF levels. AI provides a first-pass suggestion only; every level is confirmed by the editorial team against the published rubric before a course goes live.
- •Issue certificates. Certificates are triggered by learner action and gated by the credit rules, never by a model output.
- •Train on learner data. No learner name, email, attempt record or completion is sent to any model provider for training. Learner data is stored inside the EU and processed under GDPR.
- •Author courses without human sign-off. External contributors and partner organisations retain authorship; the editorial team, or the named author, reviews every module before it is published.
The editorial and AI-assurance methodology
EANE's editorial and AI-assurance methodology was developed by the Editorial Team with reference to the EU Council Recommendation on micro-credentials (2022/C 243/02) and on the European Qualifications Framework (2017/C 189/03), established learning-science literature on spaced retrieval and durable memory formation (Ebbinghaus, Bjork, Squire), and consultation with Erasmus+ partner practitioners during pilot rollouts.
A standing Editorial & Quality Assurance Panel is now being formalised, and an external academic advisor position is open, ahead of our first annual audit scheduled Q3 2027.
What happens on every module, every day
The Q3 2027 external audit is a formalisation of scrutiny that runs continuously today. Every module on EANE is subject to four checks before, during and after publication:
AI Quality Audit before publication
Every module runs through the Quality Audit AI which surfaces ambiguous questions, weak distractors, reading-level outliers and tone drift. Flags are triaged by the editorial team; no module is published while a critical flag is open.
iEQF calibration against the rubric
AI provides a first-pass iEQF suggestion for each course, scored against the published rubric; every level is then confirmed by the editorial team before the course is enabled for enrolment. The rubric is public.
Failure-rate signals surface ambiguity
Any module with an unusually high fail-rate is automatically surfaced to the editorial queue via the High-Failure Audit report, so unclear questions and unclear writing are re-worked, not left to punish learners.
Learner concerns route to the Head of Academy
Any complaint about a module, a translation or a credential is answered by the Head of Academy within five working days. Substantive concerns feed the next audit's scope.
The full iEQF rubric is published at eane.eu/iEQF/rubric.
Editorial governance today
Two named roles carry day-to-day editorial and quality responsibility while the Panel is being formalised.
Editorial direction, methodology, AI-assisted editorial workflows, and convening the Editorial & Quality Assurance Panel.
PgDip Innovation & Creativity · BA (Mod.) Hons Italian & Spanish · Community media, film & TV production, curriculum design, EU project development.
Editorial workflow, monitoring and evaluation, compliance, and continuous-improvement processes across the Academy.
MSc International Management · BBA · Quality assurance & continuous improvement, international project management, governance, monitoring & evaluation.
The Editorial & Quality Assurance Panel
The Panel is a standing body being formalised ahead of the first annual external audit. It will approve rubric updates, review AI Quality Audit outputs before publication cycles, sign off on the annual audit and adjudicate any credential-integrity concerns.
Panel members, internal
One to two additional named members with editorial, pedagogical or quality-assurance expertise. Approximate commitment: annual review and audit sign-off.
External Academic Advisor
A named external advisor from a university or recognised professional body (media studies, adult education, translation studies or quality assurance in education). Commitment: annual audit oversight and methodology sign-off.
First annual audit, Q3 2027
The first annual external audit will scope the four continuous checks above, sample the AI Quality Audit outputs, verify iEQF calibration against the rubric on a random module set, and review translation fidelity in every language then published on EANE.
Findings will be published on this page, alongside the date of the next audit. Any material change to methodology triggered by an audit is recorded in the changelog below.
Audit changelog
The first entry will be added after the Q3 2027 audit closes.
Anchored in European frameworks
- EU AI Act (Regulation 2024/1689) — EANE treats every AI use as a supporting tool for a human editorial decision. AI is not used to determine access to education, assign qualifications or judge learner responses. The provider-declared iEQF level, credit rules and pass/fail logic are all human-set.
- GDPR (Regulation 2016/679) — Learner data is stored on infrastructure inside the European Union and never sent to a model provider for training.
- Council Recommendation 2022/C 243/02 — EANE publishes exclusively in the micro-credential format defined by the Council Recommendation on a European approach to micro-credentials for lifelong learning and employability.
- Council Recommendation 2017/C 189/03 — The iEQF rubric maps each level to the descriptors of the European Qualifications Framework for lifelong learning.
Which models we use today
For full transparency, here are the specific AI systems that support the editorial workflow described above. This section is reviewed periodically; any provider or model change is recorded here and in the audit changelog below.
Translations & Quality Audit
Google Gemini 2.5 Flash, used for the multilingual translation pipeline and the pre-publication Quality Audit pass.
iEQF first-pass calibration
Google Gemini 2.5 Flash, scored against the published rubric. The editorial team confirms every course-level iEQF value before enrolment opens.
Cover imagery
Google Gemini Nano Banana (image model), used to generate module cover artwork from editorial briefs. Every cover is selected by an editor before publication.
Read-aloud
The learner's browser (Web Speech API), no external model receives module text or learner data.
Last reviewed: July 2026. Provider or version changes are noted in the audit changelog immediately below.
Questions or concerns about how EANE uses AI?
Write to the Head of Academy at hello@eane.eu. We answer every substantive enquiry within five working days.