EU AI Act Compliance

Guidance for organisations deploying or self-hosting a Qiplim Studio instance within the European Union.

Last updated: April 2026. Based on Regulation (EU) 2024/1689.

What is the AI Act?

The EU AI Act (Regulation 2024/1689) is the world's first comprehensive legal framework for artificial intelligence. It classifies AI systems by risk level and imposes obligations on providers and deployers. Key compliance deadlines run from February 2025 to August 2027.

How is Qiplim Studio classified?

Qiplim Studio uses AI (LLMs) to generate educational content: quizzes, summaries, flashcards, podcasts, etc. from documents. The classification depends on how you deploy it.

Generally NOT high-risk

Content generation (creating learning materials from documents) is not listed in Annex III. If Qiplim Studio is used as a content creation tool where educators review and approve outputs before use, it falls under limited or minimal risk.

Could become high-risk

Under Annex III, section 3 (Education and vocational training), these uses ARE high-risk:

  • Determining access or admission to educational institutions
  • Evaluating learning outcomes, especially when used to steer the learning process
  • Assessing the appropriate education level an individual will receive
  • Monitoring and detecting prohibited behaviour during tests

If your deployment uses quiz/assessment results to automatically determine grades, certifications, or student progression without human review, it may be classified as high-risk.

Your obligations by deployment scenario

Content creation tool (minimal risk)

Educators use Qiplim to generate content, review it, then deliver to students.

Transparency: inform users that content is AI-generated. No further AI Act obligations.

Self-paced learning with tracking (limited risk)

Students interact with generated content. Results are tracked but not used for decisions.

Transparency: clearly label AI-generated content. Inform students that AI is involved. Keep human oversight for any assessment decisions.

Automated assessment or progression (high-risk)

AI-generated quiz results directly influence grades, certifications, or student access.

Full high-risk compliance required: risk management system, data governance, technical documentation, transparency to deployers and users, human oversight measures, accuracy and robustness testing, EU database registration.

Compliance checklist for deployers

1

Transparency: label all AI-generated content as such

2

Human oversight: ensure a qualified person reviews AI outputs before consequential use

3

Data governance: document what data is processed, where it is stored, and retention periods

4

Provider information: document which AI providers (Mistral, OpenAI, etc.) are used and their terms

5

BYOK keys: if using Bring Your Own Key, ensure your data processing agreements cover the AI provider

6

Risk assessment: evaluate whether your specific use case falls under Annex III high-risk categories

7

Record-keeping: enable logging of AI generations for audit purposes

8

User information: inform end users that AI is used in content generation

Technical measures in Qiplim Studio

BYOK (Bring Your Own Key)

Choose your AI provider and control data processing. Keys encrypted with AES-256-GCM.

Generation logging

All AI generations are logged with provider, model, timestamps, and token usage.

Human-in-the-loop

Generated content goes through an editor before publication. No direct student-facing auto-publish.

Self-hosting

Deploy on your infrastructure. Full control over data residency and processing.

Open core (in preparation)

Inspect the code. Verify what the AI does. No black box.

Key compliance dates

February 2025: Prohibited AI practices banned

August 2025: GPAI model obligations apply

August 2026: High-risk system requirements (Annex III) enforceable

August 2027: Full enforcement for all AI systems

Resources

This page provides general guidance and does not constitute legal advice. Consult a qualified legal professional to assess your specific deployment scenario.

AI Act — Qiplim Studio