Published:
Last updated:
EU AI Act and Explainable AI
The EU AI Act classifies AI systems by their risk to society. Organisations must ensure that their AI applications — particularly in high-stakes areas such as HR or credit — are explainable and auditable.
Core Concept
Prohibition of AI posing unacceptable risk (e.g. social scoring) and strict requirements for high-risk systems. For generative AI (general-purpose AI), specific transparency obligations apply regarding training data and copyright.
Relevance
- Risk Classification: Assessment of internal AI projects prior to production rollout.
- Explainability (XAI): Use of methods that make AI decisions comprehensible to humans.
- Compliance Documentation: Building technical dossiers to fulfil accountability requirements.