Compliance Mapping — Votal Shield Guardrails

This document maps Votal Shield guardrails to industry compliance frameworks to help security, risk, and audit teams evidence control coverage.

Frameworks covered:

  • NIST AI Risk Management Framework (AI RMF 1.0)
  • OWASP Top 10 for LLM Applications
  • NIST SP 800-53 (Security and Privacy Controls)
  • EU AI Act (High-Risk AI Systems)
  • ISO/IEC 42001 (AI Management System)

1. NIST AI Risk Management Framework (AI RMF 1.0)

GOVERN Function — Organizational Controls

Control ID Requirement Votal Shield Coverage
GOVERN 1.4 Legal and regulatory requirements are understood, managed, and documented pii_detection, pii_leakage (GDPR/HIPAA/CCPA), data_access_guard, per-tenant data isolation
GOVERN 1.5 Ongoing monitoring and periodic review of risk management Telemetry to Elasticsearch SIEM, Kibana dashboards, audit logs
GOVERN 1.6 Mechanisms are in place to inventory AI systems and assess risk Tenant registry, agent_key → role mapping, guardrail registry
GOVERN 4.1 Organizational accountability structures exist rbac_guard, agent_key tracing, per-tenant isolation, audit trail
GOVERN 4.3 Organizational practices for testing and evaluation Per-guardrail latency/block rate metrics in telemetry
GOVERN 5.1 Engagement with external stakeholders Exportable telemetry events (ECS-compliant) for compliance reporting
GOVERN 6.1 High-risk AI systems managed with appropriate controls budget_controls, rate_limiter, sensitive_action_confirmation, action_guard

MAP Function — Context and Risk Identification

Control ID Requirement Votal Shield Coverage
MAP 2.3 AI system purpose and use are documented topic_enforcement (system purpose enforcement), scope_boundaries per role
MAP 3.4 Risks mapped to business context data_access_guard, action_classification, tool_allowlist per role
MAP 4.1 Impact assessment of AI system risks action_classification with max_risk_per_role
MAP 5.1 Likelihood and impact of identified risks assessed event.risk_score (0-100) computed per event, severity mapping
MAP 5.2 Risks and benefits characterized Full telemetry schema with attack_type, blocked_guardrails, risk_score

MEASURE Function — Trustworthiness Characteristics

Control ID Characteristic Votal Shield Coverage
MEASURE 2.3 Valid and Reliable safety_check, adversarial_detection, factual_grounding, hallucinated_links
MEASURE 2.5 AI system performance is evaluated Per-guardrail latency metrics, block rates via telemetry
MEASURE 2.6 Safe safety_check, keyword_blocklist, toxicity, topic_restriction, topic_enforcement
MEASURE 2.7 Secure and Resilient adversarial_detection, system_prompt_leak, regex_pattern, rate_limiter, memory_injection_detection
MEASURE 2.8 Accountable and Transparent Telemetry with trace.id + agent.key, audit_log, chain_of_thought_monitoring
MEASURE 2.9 Explainable and Interpretable guardrail_results with reasoning, telemetry events tagged with attack_type
MEASURE 2.10 Privacy-Enhanced pii_detection, pii_leakage, memory_pii_scrubbing, role_redaction
MEASURE 2.11 Fair, with Harmful Bias Managed bias_detection, tone_enforcement
MEASURE 2.13 Effectiveness Per-guardrail metrics via telemetry (latency, block rate, false positive tracking)
MEASURE 3.1 Approaches identified for TEVV Guardrail pipeline with tiered execution (fast/medium/slow)
MEASURE 4.2 Measurement results are captured Elasticsearch SIEM stores all events with ECS-compliant schema

MANAGE Function — Risk Response

Control ID Requirement Votal Shield Coverage
MANAGE 1.3 AI risks are treated based on priority Guardrail action field (block/warn/log) + event.risk_score tiering
MANAGE 2.3 Mechanisms exist to sustain AI system value Runtime config updates via /v1/shield/config, /v1/admin/tenants
MANAGE 2.4 Mechanisms to supersede, disengage, or deactivate AI sensitive_action_confirmation, action_guard, rbac_guard block capabilities
MANAGE 3.1 AI risks and benefits from third-party resources mcp_guard with trust scoring, tool_allowlist
MANAGE 4.1 Post-deployment AI monitoring Full telemetry pipeline to Elasticsearch/SIEM, chain_of_thought_monitoring, loop_detection
MANAGE 4.2 Measurable continuous improvement Audit logs, per-guardrail metrics, Kibana dashboards
MANAGE 4.3 Incidents and errors communicated ES alerts on event.kind: "alert" + event.severity: "critical"

2. OWASP Top 10 for LLM Applications

OWASP ID Risk Votal Shield Guardrails
LLM01 Prompt Injection adversarial_detection, memory_injection_detection, system_prompt_leak, chain_of_thought_monitoring
LLM02 Insecure Output Handling tool_output_sanitization, pii_leakage, role_redaction, regex_pattern
LLM03 Training Data Poisoning (out of scope — requires model training controls)
LLM04 Model Denial of Service rate_limiter, length_limit, budget_controls, tool_call_rate_limiting, loop_detection
LLM05 Supply Chain Vulnerabilities mcp_guard with MCP server trust scoring
LLM06 Sensitive Information Disclosure pii_detection, pii_leakage, memory_pii_scrubbing, data_access_guard, system_prompt_leak
LLM07 Insecure Plugin Design tool_allowlist, tool_use_control, tool_call_validation, mcp_guard
LLM08 Excessive Agency action_guard, sensitive_action_confirmation, scope_boundaries, delegation_control
LLM09 Overreliance factual_grounding, hallucinated_links, chain_of_thought_monitoring
LLM10 Model Theft rate_limiter, rbac_guard, audit_log, budget_controls

3. NIST SP 800-53 Rev 5 (Security and Privacy Controls)

Access Control (AC)

Control ID Requirement Votal Shield Coverage
AC-2 Account Management rbac_guard, tenant API key management via /v1/admin/tenants
AC-3 Access Enforcement rbac_guard, scope_boundaries, data_access_guard
AC-4 Information Flow Enforcement data_access_guard, scope_boundaries, per-role resource access
AC-5 Separation of Duties rbac_guard with per-role tool/data restrictions
AC-6 Least Privilege tool_allowlist, action_guard, per-role data clearance
AC-7 Unsuccessful Logon Attempts rate_limiter (per-agent rate limiting)

Audit and Accountability (AU)

Control ID Requirement Votal Shield Coverage
AU-2 Event Logging TelemetryMiddleware → Elasticsearch SIEM
AU-3 Content of Audit Records Full event schema (ECS-compliant) with timestamp, source, action, outcome
AU-6 Audit Record Review, Analysis, and Reporting Kibana dashboards, alerts on high-risk events
AU-9 Protection of Audit Information SIEM with API key auth, immutable log storage
AU-12 Audit Record Generation audit_log SQLite + ES exporter

Identification and Authentication (IA)

Control ID Requirement Votal Shield Coverage
IA-2 Identification and Authentication API key auth (AuthMiddleware), agent_key identity
IA-5 Authenticator Management SHA-256 hashed API keys, rotation via env vars
IA-8 Identification and Authentication (non-organizational) Per-tenant API keys, tenant-scoped identities

System and Communications Protection (SC)

Control ID Requirement Votal Shield Coverage
SC-7 Boundary Protection scope_boundaries, rbac_guard, multi-tenant isolation
SC-8 Transmission Confidentiality and Integrity HTTPS required for telemetry exporters (ES, Splunk, OTLP)
SC-28 Protection of Information at Rest memory_pii_scrubbing, role_redaction

System and Information Integrity (SI)

Control ID Requirement Votal Shield Coverage
SI-3 Malicious Code Protection keyword_blocklist, regex_pattern, adversarial_detection
SI-4 System Monitoring chain_of_thought_monitoring, context_window_guardrails, telemetry pipeline
SI-7 Software, Firmware, Information Integrity tool_call_validation, memory_guardrails
SI-10 Information Input Validation regex_pattern, length_limit, keyword_blocklist, tool_call_validation
SI-11 Error Handling Graceful guardrail failure handling (pipeline catches exceptions per guardrail)

4. EU AI Act (High-Risk AI Systems)

Article Requirement Votal Shield Coverage
Art. 9 Risk Management System Full guardrail pipeline + SIEM telemetry + audit logs
Art. 10 Data and Data Governance pii_detection, data_access_guard, memory_pii_scrubbing
Art. 12 Record-Keeping Audit logs (SQLite) + ES telemetry (immutable event log)
Art. 13 Transparency and Provision of Information guardrail_results in API responses with reasoning
Art. 14 Human Oversight sensitive_action_confirmation, action_guard, workflow approval gates
Art. 15 Accuracy, Robustness, and Cybersecurity adversarial_detection, rate_limiter, budget_controls, regex_pattern
Art. 17 Quality Management System Per-guardrail metrics, continuous monitoring via SIEM
Art. 61 Post-Market Monitoring Telemetry pipeline, incident alerts, Kibana dashboards

5. ISO/IEC 42001 (AI Management System)

Clause Requirement Votal Shield Coverage
6.1.2 AI risk assessment event.risk_score, per-guardrail risk tiers
6.1.3 AI risk treatment Guardrail action field (block/warn/log), runtime config updates
8.2 AI system impact assessment action_classification with risk scoring per role
8.3 AI system design and development Guardrail pipeline with three-tier execution (fast/medium/slow)
8.4 Third-party relationships mcp_guard, tool_allowlist, per-tenant isolation
9.1 Monitoring, measurement, analysis, evaluation Full telemetry pipeline to SIEM
9.2 Internal audit audit_log with query API
10 Improvement Runtime guardrail config updates, continuous metrics

6. Guardrail → Control Reverse Index

Quick lookup: given a guardrail, what controls does it cover?

Guardrail NIST AI RMF OWASP LLM NIST 800-53
keyword_blocklist MEASURE 2.6 LLM06 SI-3, SI-10
length_limit MEASURE 2.7 LLM04 SI-10
regex_pattern MEASURE 2.7 LLM01, LLM06 SI-10
pii_detection MEASURE 2.10, GOVERN 1.4 LLM06 SC-28
rate_limiter MEASURE 2.7 LLM04, LLM10 AC-7
system_prompt_leak MEASURE 2.7 LLM01, LLM06 SI-4
toxicity MEASURE 2.6 SI-3
safety_check MEASURE 2.3, 2.6 LLM01 SI-4
adversarial_detection MEASURE 2.7 LLM01 SI-3, SI-4
topic_restriction MEASURE 2.6 AC-4
topic_enforcement MAP 2.3 AC-4
role_redaction MEASURE 2.10 LLM02 SC-28
hallucinated_links MEASURE 2.3 LLM09
tone_enforcement MEASURE 2.11
bias_detection MEASURE 2.11
pii_leakage MEASURE 2.10, GOVERN 1.4 LLM02, LLM06 SC-28
competitor_mention GOVERN 5.1
rbac_guard GOVERN 4.1 LLM08 AC-2, AC-3, AC-6
data_access_guard GOVERN 1.4, MAP 3.4 LLM06 AC-3, AC-4
mcp_guard MANAGE 3.1 LLM05, LLM07 SC-7
action_guard MANAGE 2.4 LLM08 AC-6
tool_allowlist MAP 3.4 LLM07 AC-6
tool_use_control MAP 3.4 LLM07 AC-3
tool_call_rate_limiting MEASURE 2.7 LLM04 AC-7
tool_call_validation MEASURE 2.7 LLM07 SI-7, SI-10
tool_output_sanitization MEASURE 2.10 LLM02 SC-28
sensitive_action_confirmation MANAGE 2.4 LLM08 AC-6
action_classification MAP 3.4, MAP 5.1 LLM08 AC-6
scope_boundaries GOVERN 4.1 LLM08 AC-3, AC-4, SC-7
loop_detection MEASURE 2.7 LLM04 SI-4
budget_controls GOVERN 6.1 LLM04, LLM10
delegation_control GOVERN 4.1 LLM08 AC-5, AC-6
memory_guardrails MEASURE 2.7 SI-7
memory_pii_scrubbing MEASURE 2.10 LLM06 SC-28
memory_injection_detection MEASURE 2.7 LLM01 SI-3, SI-4
memory_retention_policies GOVERN 1.4 AU-11
memory_access_control GOVERN 4.1 LLM06 AC-3
chain_of_thought_monitoring MEASURE 2.8 LLM01, LLM09 SI-4
context_window_guardrails MEASURE 2.7 LLM04 SI-4

7. Evidencing Compliance

For auditors asking “how do you demonstrate control X?”, point to:

Evidence Type Location
Guardrail config config/default.yaml + per-tenant configs in Redis
Event logs Elasticsearch index votal-shield-logs*
Audit trail SQLite audit.db (queryable via /v1/shield/audit)
Per-guardrail metrics votal.guardrail.latency_ms, votal.guardrail.passed in ES
Risk scoring event.risk_score, event.severity in every event
Trace correlation trace.id, agent.key, votal.session_id across all events
Blocked attack samples Query ES for event.kind: "alert" AND event.risk_score >= 90
RBAC enforcement rbac_guard decisions logged with agent.key + votal.role_name

8. Kibana Queries for Compliance Reports

# LLM01 (Prompt Injection) — all blocked attacks
votal.attack_type: ("prompt_injection" OR "jailbreak") AND event.outcome: "failure"

# LLM06 (Sensitive Info Disclosure) — all PII events
votal.guardrail.name: ("pii_detection" OR "pii_leakage") AND votal.guardrail.passed: false

# MEASURE 2.10 (Privacy) — PII blocks by tenant
agent.key: acme-* AND votal.guardrail.name: pii_*

# MANAGE 4.3 (Incidents) — critical alerts last 24h
event.kind: "alert" AND event.severity: "critical" AND @timestamp >= "now-24h"

# AC-2 (Account Management) — new agent activity
agent.key: * AND event.action: "request"

# SI-4 (Monitoring) — all guardrail activity
NOT url.path: ("/health" OR "/ping")

References