EU AI Act Compliance¶

Kiket provides comprehensive support for EU AI Act Article 12 compliance through blockchain-verified audit logging of AI operations.

Overview¶

The EU AI Act requires organizations using AI systems to maintain detailed logs of:

AI system operations and decisions
Human oversight activities
Risk assessments and mitigations
Transparency measures

Kiket's blockchain audit trail provides tamper-proof evidence of compliance.

Article 12 Requirements¶

Automatic Logging¶

The EU AI Act mandates automatic logging of AI system operations. Kiket captures:

Requirement	Kiket Implementation
Operation timestamps	ISO 8601 timestamps for all AI invocations
Input/output data references	Content hashes of inputs and outputs
Decision outcomes	AI recommendations and actions taken
Human interventions	Override and approval records
System version	AI agent version tracked per operation

Audit Trail Integrity¶

All AI audit logs are:

Cryptographically hashed - SHA-256 content hashes
Merkle tree structured - Efficient verification
Blockchain anchored - Polygon mainnet immutability
Independently verifiable - Third-party verification possible

AI Operations Logging¶

What Gets Logged¶

Kiket automatically logs all AI agent operations:

AI Operation Audit Log
├── Operation type (assignment, suggestion, analysis, etc.)
├── AI agent identifier and version
├── Input context (hashed)
├── Output/recommendation
├── Decision status (accepted, rejected, modified)
├── Human reviewer (if applicable)
├── Token usage and cost
├── Latency metrics
└── Blockchain anchor reference

Example Audit Entry¶

{
  "operation": "ai_assignment_suggested",
  "agent_id": "assignment-agent-v2",
  "status": "accepted",
  "metadata": {
    "issue_id": 123,
    "suggested_assignee": "user_456",
    "confidence_score": 0.92,
    "reasoning_summary": "Skill match and availability"
  },
  "user_id": 789,
  "occurred_at": "2026-01-15T10:30:00Z",
  "blockchain_anchored_at": "2026-01-15T11:00:00Z"
}

Human Oversight Compliance¶

Override Tracking¶

When users override AI decisions:

Original AI recommendation recorded
Override action and reason captured
Both entries blockchain-anchored
Audit trail shows full decision history

Approval Workflows¶

For AI actions requiring human approval:

Approval request logged
Reviewer identity captured
Decision (approve/reject) recorded
Time-to-decision tracked

Generating EU AI Act Reports¶

Via Web Interface¶

Navigate to Blockchain > Dashboard
Click Compliance Reports
Select EU AI Act Report
Choose date range (default: 6 months)
Click Generate PDF

Via API¶

curl -X GET "https://app.kiket.dev/api/v1/audit/reports/eu_ai_act.pdf" \
  -H "Authorization: Bearer <token>" \
  -o eu_ai_act_report.pdf

Via CLI¶

kiket audit export --type eu_ai_act --output eu_ai_act_report.pdf

Report Contents¶

The EU AI Act compliance report includes:

1. AI System Inventory¶

List of all AI agents in use
Version information
Deployment dates
Risk classification (where applicable)

2. Operation Statistics¶

Metric	Description
Total Operations	Number of AI invocations
Unique Users	Users interacting with AI
Acceptance Rate	% of AI recommendations accepted
Override Rate	% of AI recommendations overridden
Average Response Time	AI processing latency

3. Human Oversight Summary¶

Approval workflow statistics
Override patterns and reasons
Reviewer distribution
Response time metrics

4. Blockchain Verification¶

Anchoring coverage (% of operations anchored)
Anchor transaction references
Verification instructions
QR codes for key transactions

5. Risk Assessment Summary¶

Identified risk factors
Mitigation measures in place
Compliance status indicators

Configuration¶

AI Agent Registration¶

Each AI agent should be registered with:

# .kiket/agents/assignment-agent.yaml
manifest_version: 1
agent:
  name: Assignment Agent
  version: 2.0.0
  risk_classification: limited  # minimal, limited, high
  purpose: "Suggest optimal issue assignees based on skills and availability"
  oversight_required: false  # Set true for high-risk

Immediate Anchoring¶

For compliance-critical AI operations, enable immediate anchoring:

# Organization settings
ai_audit:
  immediate_anchor_operations:
    - ai_assignment_accepted
    - ai_recommendation_overridden
    - ai_high_risk_decision

Best Practices¶

Regular Reporting¶

Generate monthly EU AI Act reports
Review AI acceptance/override patterns
Document any compliance gaps

Risk Assessment¶

Classify AI agents by risk level
Enable appropriate oversight for high-risk systems
Document mitigation measures

Documentation¶

Maintain AI system documentation
Keep records of human oversight procedures
Document compliance review processes

Compliance Checklist¶

Requirement	Implementation	Status
Automatic logging	AI audit logs	Auto
Tamper-proof storage	Blockchain anchoring	Auto
Human oversight records	Override/approval tracking	Auto
Regular reporting	Monthly PDF exports	Manual
Risk assessment	Agent classification	Manual
Documentation	System inventory	Manual

EU AI Act Official Text
[Blockchain Audit Trail(../features/blockchain-audit-trail.md)
[Compliance Reports(blockchain-reports.md)
[AI Agents Documentation(../platform/ai-agents.md)