the EU keeps publishing the memo. it still hasn't built the credential.
Europe has the frameworks and four fresh sets of guidelines. What it still doesn't have, in operational form, is the credential in the middle of the stack — independently assessed, portable across all 27 states, the thing a vendor course maps to rather than competes with.
On 9 June 2026 the European Commission did the thing it does best. It updated a document. The ethical guidelines on the use of AI and data in teaching and learning — first issued in 2022, now revised — landed in English with translations to follow, prepared by a working group convened through the European Digital Education Hub.
The update was prepared by the Working Group on the Ethical Use of AI and Data in Education, convened through the European Digital Education Hub, and was last updated 9 June 2026.
I've read it. It's a good document — guiding questions, classroom scenarios, a glossary.
It includes guiding questions and classroom scenarios, explanations of key principles behind responsible AI use, and an updated glossary defining AI and data-related terms.
And here is the uncomfortable part, said plainly: it is non-binding, aimed mostly at primary and secondary teachers, and it does not move a single one of the numbers Europe is actually failing on.
These guidelines are non-binding and are intended to support teachers, school leaders and education authorities.
I've spent thirty years watching institutions confuse publishing a framework with changing an outcome. The two are not the same. They are barely related.
the number that matters, and the one nobody hits
Start with the target everyone cites and nobody is on track for.
The EU aims to achieve a digitally skilled population by 2030, where at least 80% of those aged 16-74 have at least basic digital skills.
Where are we?
Only 55.6% of the EU's population have at least basic digital skills.
That's not a rounding error from target. That's twenty-four points of daylight.
The professional side is worse in absolute terms.
At least 20 million ICT specialists should be employed in the EU by 2030; in 2024 there were an estimated 10.3 million, accounting for 5% of the workforce.
You need to roughly double the specialist workforce in under four years. One independent analysis put it bluntly:
at the current pace, the EU would need to increase digital skills adoption nearly nine times faster to meet its 80% target by 2030.
Take the exact multiple with a pinch of salt — it depends on which baseline year you anchor to — but the direction is not in dispute. Ten member states are reportedly moving backwards.
So when I see another set of ethical guidelines, however competent, my first question as a practitioner is: which of these two curves does it bend? Answer: neither directly. Guidance is necessary. It is nowhere near sufficient.
the credentialing question, which is the real one
Here's where it gets interesting. The bottleneck isn't awareness.
In 2024, 37% of lower secondary teachers reported using AI in their work, according to OECD's TALIS survey.
People are already using the tools.
In recent surveys, a large majority — 87% — believe that all teachers should be equipped with the skills to use and understand AI.
The demand signal is loud. What's missing is a portable, trusted way to certify that someone actually has the skill.
Europe has been circling this for years. The European Digital Skills Certificate was scoped back in 2021, built on DigComp, with a feasibility study and a pilot.
The EDSC is built upon the European Digital Competence Framework (DigComp), and in September 2022 a feasibility study was launched to explore its feasibility.
Four years on, the operational reality that has actually shipped is the adjacent infrastructure: European Digital Credentials for Learning — tamper-evident, electronically sealed records.
A European Digital Credential for Learning is a verifiable digital version of a credential issued to a learner; these include diplomas, training certificates, micro-credentials, and certificates of participation, signed with an electronic seal.
Quality and Qualifications Ireland was running launch sessions on it as recently as mid-May 2026. Useful plumbing. But plumbing is not a standard for what AI competence means — it's a way to sign whatever you decide it means.
Into that vacuum walk the vendors. In March 2026 Google launched AI Works for Europe and told us it would bring its professional certificate to the continent.
At the Future of Work Forum in Riga, Google launched AI Works for Europe, with a first commitment including $30 million of additional support for Google.org's European AI Opportunity Fund and the Google AI Professional Certificate available in ten European languages.
OpenAI went bigger in its EU blueprint, proposing to
train 100 million Europeans in foundational AI skills by 2030 through freely accessible online courses in all official EU languages.
I will say what a board would want me to say about that last one. A hundred million is a press-release number. Attendance is not attainment, and a completion certificate from a model vendor is a marketing asset before it's a labour-market signal. If "100 million AI citizens" means 100 million people who watched some videos, it will move the curves above by approximately nothing. I'd bet against it meaning anything unless it's tied to an independent assessment standard Europe owns — and right now Europe doesn't have one that's operational at scale.
what a real curriculum looks like — and why it's hard
I have a stake here, so I'll declare it. My firm, Real AI, is one of the four SME partners in PANORAIMA, and I was a founding contributor to its predecessor, HCAIM. Read the rest with that in mind.
HCAIM — the Human-Centred AI Master's — was the serious attempt at the degree-level end.
Launched in 2021, it ended in 2024; a consortium of universities, excellence centres and companies from five EU countries developed a 60-credit English-language master's curriculum, run at BME from 2022, mainly for ICT students.
Its successor, PANORAIMA, pushes the same human-centred content out beyond computer science.
PANORAIMA extends AI education to healthcare, media, law, management and finance professionals, bringing together 16 consortium members — 8 European universities, 4 research centres and 4 SMEs — to develop AI curricula aligned with market needs.
The pilot specialisation tracks are due to begin this autumn.
First pilot runs of the developed specialisation tracks start in September 2026, with full availability including online modules from September 2027.
Now the part the brochures skip. HCAIM's honest internal accounting reported something most EU programmes won't put in writing:
about 90 BME students joined the programme since inception, with three qualifying in 2022, eight in 2023 and five in spring 2024 — and a relatively high dropout rate of around 50%, as students had to acquire extra credits that often didn't fit their two-year master's.
Hold that next to the 100-million ambition. A rigorous, well-funded, multi-university programme graduated people in the low dozens and lost half its starters to structural friction. That is the actual difficulty of producing certified AI competence. It is slow, expensive, and it does not scale by decree.
That's not an argument against PANORAIMA — I wouldn't have put Real AI's name on it if it were. It's an argument for honesty about the gradient. The modular, self-standing upskilling units PANORAIMA is building matter precisely because the full master's is too heavy for a working radiographer or compliance officer to carry.
the field note
The Council, to its credit, framed the politics correctly in May.
Its conclusions call for an ethical, safe and human-centred approach to AI in education focused on strengthening digital skills and AI literacy, guaranteeing inclusion and fairness, and empowering teachers.
Agreed, all of it. But "empowering teachers" is a budget line and an assessment standard, not a value statement.
If I were advising the team preparing the State of the Digital Decade Report 2026 — and they're explicitly stress-testing interim findings against the competitiveness and sovereignty agenda right now — I'd push one priority above the rest: finish the European AI-skills credential, make it independently assessed, make it portable across all 27 states, and make it the thing a Google or an OpenAI course maps to rather than competes with. Sovereignty in AI education isn't owning the courseware. It's owning the standard that says what counts.
Europe has the frameworks. It has the guidelines — four fresh sets of them.
In a single release the Commission put out four sets of guidelines — two new, two updated — covering ethical AI use, digital literacy, quality content selection, and teaching informatics.
What it still doesn't have, in operational form, is the credential in the middle of that stack. Until it does, the 80% target stays twenty-four points away and the 100-million headline stays a headline.
The memo is written. Build the credential.
Tarry Singh is the founder and CEO of Real AI (realai.eu), an enterprise AI advisory and deployment firm working with global enterprises on production agent systems, model risk, and AI sovereignty strategy. He also leads Earthscan (earthscan.io) for Energy AI, and is a founding contributor to the EU-funded HCAIM and PANORAIMA programmes for responsible AI education across European universities. He writes at tarrysingh.com.