What Are Policies?
Policies are the rules that define what AI requests need human review. Velatir uses advanced AI to automatically evaluate every request against your active policies, determining compliance status, risk levels, and whether human intervention is required.How Policy Evaluation Works
When an AI request comes in, Velatir’s policy engine:- Analyzes the Request - Examines function name, arguments, AI explanation, and metadata
- Applies All Active Policies - Runs the request through every enabled policy
- Generates Assessments - Creates detailed evaluation for each policy
- Determines Actions - Decides auto-approval vs. human review based on results
Policy Assessment Output
For each policy, the AI evaluator returns:Compliance Status
Compliant: Request meets policy requirements
Non-Compliant: Request violates policy or needs review
Non-Compliant: Request violates policy or needs review
Risk Level
Low: Minimal risk, standard logging
Medium: Some risk, may need disclaimers
High: Significant risk, human oversight required
Critical: Major violation, should be blocked
Medium: Some risk, may need disclaimers
High: Significant risk, human oversight required
Critical: Major violation, should be blocked
Confidence Score
0.0 - 1.0: How confident the AI is in its assessment
Higher scores mean more certain evaluations
Higher scores mean more certain evaluations
Recommendation
Auto-Approve: Safe to proceed automatically
Human Intervention Required: Needs human approval
Human Intervention Required: Needs human approval
Assessment Details
Every policy evaluation includes:- Reason - Single clear sentence explaining the decision
- Tags - Categorization labels (e.g., #PersonalData, #FinancialData)
- Policy Version - Exact version used for the assessment
- Evaluation Timestamp - When the assessment was performed
Built-in Policy Types
GDPR Compliance
GDPR Compliance
Detects processing of personal data, special categories, third-country transfers, and automated decision-making scenarios that require human oversight.
EU AI Act Compliance
EU AI Act Compliance
Identifies high-risk AI use cases under EU AI Act framework and ensures appropriate transparency and risk mitigation measures.
Bias & Fairness
Bias & Fairness
Flags interactions involving profiling, discrimination potential, or decisions based on protected attributes.
Custom Policies
Custom Policies
Organization-specific rules tailored to your industry, use cases, and compliance requirements.