Zum Inhalt springen
>_<
AI EngineeringWiki

EU AI Act

Compliance · 6 min

The EU AI Act (Regulation 2024/1689) was published on 12 July 2024 and entered into force on 1 August 2024. It is the world's first comprehensive AI regulation. Some obligations — such as the prohibitions (Art. 5) and AI literacy (Art. 4) — have been applicable since 2 February 2025. Full application with enforcement and penalties starts from 2 August 2026.

Art. 4 AI Literacy ALREADY APPLIES

The AI literacy obligation (Art. 4) has been in force since 2 February 2025. There is no single certification, no mandatory AI Officer role, and no one-size-fits-all training requirement. Enforcement and supervision starts from August 2026.

Risk Classes

ClassExamplesRequirements
UnacceptableSocial scoring, manipulationBANNED
HighHR, credit, biometricFull compliance
LimitedChatbots, deepfakesTransparency
MinimalSpam filtersNone

Typical Business AI Scenarios

ScenarioRisk LevelObligations
LLM for content creationMinimal to LimitedTransparency when publishing without human review
AI customer support chatbotLimitedUsers must know they are talking to AI. If it makes service decisions, may be high-risk.
Internal productivity (email, code, summaries)MinimalNo specific obligations, but documentation recommended
AI for recruiting (CV screening, candidate ranking)HighFull compliance: documentation, risk management, human oversight

High-Risk Requirements Detail

  • Risk management system: Documented, ongoing process to identify, analyze, and mitigate risks — not a one-time assessment
  • Data governance: Training and test data must meet quality criteria. Document data sources, preparation methods, and potential biases
  • Technical documentation: Detailed documentation before market placement including intended purpose, design methodology, and test results
  • Record-keeping: Automatic logging of AI system operations for traceability. Provide to authorities on request
  • Transparency: Clear information about capabilities, limitations, and intended use
  • Human oversight: Humans can understand capabilities and limits, intervene in decisions, and choose not to use the system

Timeline

  • 12 Jul 2024: Published in EU Official Journal
  • 1 Aug 2024: Entered into force
  • 2 Feb 2025: Prohibitions (Art. 5) + AI Literacy (Art. 4) ALREADY APPLY
  • 2 Aug 2025: Governance rules + GPAI models
  • 2 Aug 2026: Full application (high-risk etc.) + enforcement/supervision starts
  • 2 Aug 2027: High-risk AI in regulated products

Penalties

  • Unacceptable risk: EUR 35M or 7% global revenue
  • Non-compliance: EUR 15M or 3% global revenue
  • Incorrect information: EUR 7.5M or 1% global revenue

What Local AI Helps With

  • No third-country transfer
  • Full documentation possible
  • Transparency easier to implement
  • Data stays in EU

Sources

Next step: operationalize compliance

Use ready-to-run GDPR templates, checklists and practical guidance for AI systems that need documentation and auditability.

Why AI Engineering
  • Local and self-hosted by default
  • Documented and auditable
  • Built from our own runtime
  • Made in Austria
Not legal advice.