AI Layer: AI Incidents, AI Register and Model Cards
EterShield includes a set of modules dedicated to the governance of Artificial Intelligence systems.
AI Incidents
Specialised record of incidents involving AI systems (e.g. hallucinations with impact, incorrect automated decisions, data leakage via prompt, misuse of models). It follows the same severity structure and lifecycle as general incidents (section 10), with additional AI-specific fields (model involved, type of failure, training/inference data affected).
AI Register / Model Cards
Inventory of all AI models and systems used by the organisation (in-house or third-party). For each model, the system automatically generates a Model Card, a structured document with:
- Purpose and scope of use
- Training data and known limitations
- Risk level (aligned with the EU AI Act risk classification)
- Mitigation measures implemented
Model Cards can be exported as Markdown via api/reports/model-card-md.
Automatic AI classification
The AI supports automatic classification of content and risk across several platform modules (e.g. categorisation of Knowledge Base items, section 3.2).