EU AI Act compliance
AI Act compliance, by design.
The EU AI Act is in full force in 2026. Here's how Omnilink AI complies — and helps clients comply too.
Our position in 4 points
- →1. We classify every AI system we deploy by risk level (minimal / limited / high / unacceptable). We refuse to build unacceptable-risk systems, period.
- →2. We maintain a registry of all AI systems in operation, with model card, data sources, limitations and known biases.
- →3. We perform DPIA (and FRIA when high-risk) before any system goes live, with formal client signoff.
- →4. Our chatbot, content generators and lead scorers all expose a clear AI disclosure. No deceptive AI use.
Registry of our active AI systems
Every AI system used on this website is documented below. Updated continuously as required by AI Act art. 12.
Omnilink AI Chatbot
Limited risk- Model
- Claude Sonnet 4.6 (Anthropic)
- Purpose
- Assist visitors browsing the catalogue, qualify intent, route to human contact.
- Data handling
- No personal data sent to model. Conversation history persisted in EU MySQL DB.
- Limitations
- May give imprecise answers. User informed via disclosure in UI.
AI Maturity Assessment Scorer
Minimal risk- Model
- Custom rule-based scoring (no LLM)
- Purpose
- Score AI maturity 0-100 based on 30-question questionnaire.
- Data handling
- User answers and personal data stored in EU MySQL.
- Limitations
- Indicative result. Manual consultation recommended for final decisions.
Lead Scoring (BANT light)
Minimal risk- Model
- Custom heuristics on form data (no LLM)
- Purpose
- Internal lead prioritization for outreach SLA.
- Data handling
- Form data only.
- Limitations
- No automated decision making affecting users — only internal triage.
Documents available on request
- 📄 DPIA template (GDPR-ready)
- 📄 FRIA template (high-risk AI Act)
- 📄 AI System registry (Excel + JSON)
- 📄 DPA standard for AI processing
- 📄 Security questionnaire pre-compiled
Email dpo@omnilink.it to request any of the above.