| name | eu-ai-act-tiering |
| description | Classifies proposed AI deployments into strict EU AI Act risk tiers. |
| version | 1.0.0 |
Objective
Analyze the system description and assign an irreversible risk classification based on the EU AI Act legal framework.
Execution Protocol
- Unacceptable Risk Check: Does the system use subliminal techniques, exploit vulnerabilities, biometric categorization by sensitive traits, or social scoring? (If yes -> UNACCEPTABLE).
- High Risk Check: Is the system used in critical infrastructure, employment/HR, essential private/public services (e.g., healthcare triage, insurance eligibility), or law enforcement? (If yes -> HIGH RISK).
- Transparency Risk Check: Is the system an AI chatbot, deepfake generator, or emotion recognition system? (If yes -> LIMITED RISK, requires transparency labeling).
- Minimal Risk: All other applications (e.g., spam filters, inventory optimization).
Constraints
- An insurance claims processing or health triage AI is automatically HIGH RISK.
- Provide the tier and exactly one sentence of justification citing the specific use-case trigger.
Output Format
Risk Classification: [TIER]
Justification: [Reasoning]