GDPR (minors)
Regulation (EU) 2016/679
RequiresMinors’ data carries reinforced consent and minimization duties.
HelmcodeZero logs and EU-only inference; on-premise so student data stays inside the institution.
industries · EdTech
Educational AI where the student’s data is not the product
Reinforced GDPR for minors, high-risk rules where AI evaluates learning, and sovereignty for public universities: educational AI on EU infrastructure, on-premise if needed.
compliance
Compliance, built into the stack.
Every regulatory demand mapped to a platform capability that ships built in, with nothing to configure.
AI Act
Regulation (EU) 2024/1689, Annex III 3 · deadline under revision
RequiresAI that determines access to education or evaluates learning outcomes is high-risk: traceability and human oversight.
HelmcodeOpen, documentable models that support the teacher’s decision instead of replacing it.
Sovereignty
EU infrastructure, public universities
RequiresPublic universities should keep educational data under European control.
HelmcodeOpen weights on EU infrastructure or on-premise, no foreign hyperscaler in the loop.
This page is an informational overview, not legal advice. For your obligations and the risk classification of each system, consult qualified legal counsel. AI Act Guide →
what the law names in education
What the AI Act says about a classroom.
Education gets its own entry in the AI Act, and it is unusually specific: four named uses in the high-risk annex, one practice banned outright, and a right the student can exercise. Worth reading before anyone pitches you a tool.
01
Admission and assessment are named
Annex III point 3 covers deciding access or admission and assigning people to institutions, evaluating learning outcomes including when they steer someone’s learning path, and assessing the level of education a person will receive or be able to access, at all levels.
02
So is exam proctoring
Point 3(d) names monitoring and detecting prohibited behaviour of students during tests. Remote proctoring is not a grey area you can argue your way out of: it is written into the high-risk list, in those words.
03
Reading emotions is banned, not regulated
Article 5(1)(f) prohibits AI systems that infer emotions of a person in the workplace and in education institutions, save for medical or safety reasons. This is the prohibition list, not the high-risk list, and prohibitions have applied since February 2025.
04
The student can ask why
Article 86 gives the right to a clear and meaningful explanation of the system’s role in a decision, and it excludes only point 2 of Annex III, so education is covered. You can only write that explanation about a system whose behaviour you can examine.
EU AI Act · Regulation (EU) 2024/1689 Regulation (EU) 2024/1689 of 13 June 2024: Annex III point 3, Article 5(1)(f) and Article 86. Prohibited practices have applied since 2 February 2025. The start date for the Annex III high-risk obligations is being amended, with a postponement to 2 December 2027 agreed by the co-legislators in May 2026 and pending formal adoption as of July 2026. read the report →
use cases
Your most common use cases.
The cases with the most traction in the sector, each with its own page in detail.
RAG over internal knowledge Private tutors and RAG over course materials, grounded and cited. Semantic search & recommendation Semantic search across syllabi and educational resources. Professional copilots Assistants for teachers and administration, under the institution’s control. Voice & transcription Class transcription for accessibility, processed privately. Research Guided research for students and staff, with sources cited so the work can be checked rather than trusted. Anonymisation Student data stripped before a dataset is used for analysis or research, which minors make non-negotiable.
Recommended open models.
A starting point per task type. The full guide maps 80 cases to the open model for each one.
- DeepSeek V4 Flash in Helmcode
- Tutors and RAG over syllabi at flat cost.
- qwen3-embedding + rerank in Helmcode
- Semantic search over educational materials.
- Whisper large-v3 in Helmcode
- Class transcription for accessibility, on your own infrastructure.
We are distilling and quantizing these open models into small, tightly specialised versions, trained for one task rather than for all of them. A model like that runs on less hardware, answers faster and fits where the big one does not, your own datacenter included. If you have a process with volume and stable criteria, that is the conversation we want to have with you.
// faq
Questions, answered.
What the sector's technical, compliance and business teams ask.
How do you protect minors’ data?
Prompts are never stored (zero logs) and inference runs only on EU infrastructure; on-premise keeps student data inside the institution, which matters for the reinforced GDPR duties around minors.
Is AI that evaluates students high-risk?
AI that determines access to education or evaluates learning outcomes is high-risk under the AI Act (Regulation (EU) 2024/1689, Annex III point 3). The start date moved: a May 2026 agreement takes the Annex III obligations to December 2027, pending formal adoption as of July 2026. The prohibition on inferring emotions in education, by contrast, has applied since February 2025. An open, documentable stack that supports rather than replaces the teacher makes compliance easier; the classification of each system is for your legal team.
Is it a fit for public universities?
Yes. Open weights on EU infrastructure or on-premise keep educational data under European control, with no foreign hyperscaler in the loop, the usual requirement for public institutions.
What is the entry use case?
A private tutor or a RAG assistant over course materials: high value, and the student data never leaves the institution.
// get started
START BURNING TOKENS
Skip the AI infra work. Deploy your first private inference endpoint today.
Flat rate. EU data. OpenAI API compatible.