Critical Infrastructure Use Case

Organizations in energy, water, transportation, and telecommunications deploying AI must protect operational technology (OT) environments and comply with NERC CIP, TSA directives, and sector-specific cybersecurity requirements.

Use this page when

• You are deploying AI in energy, water, transportation, or telecommunications environments subject to NERC CIP, TSA directives, or NIST CSF.
• You need to protect operational technology (OT) environments from AI-generated commands and enforce ICS/SCADA data isolation.
• You want human-in-the-loop controls for safety-critical AI decisions affecting critical systems.

Primary audience

• Primary: Technical Leaders
• Secondary: Technical Engineers, AI Agents

Regulatory Requirements

Regulation	Requirement	Keeptrusts Policy
NERC CIP	Bulk electric system protection	dlp-filter, rbac
TSA SD-02	Pipeline cybersecurity	cjis-mode, audit-logger
NIST CSF	Cybersecurity framework	rbac, dlp-filter, audit-logger
ICS-CERT	Industrial control system security	dlp-filter, safety-filter

Complete Policy Configuration

pack:
  name: critical-infrastructure-governance
  version: 1.0.0
  enabled: true
policies:
  chain:
  - prompt-injection
  - rbac
  - cjis-mode
  - dlp-filter
  - safety-filter
  - pii-detector
  - human-oversight
  - quality-scorer
  - audit-logger
policy:
  prompt-injection: {}
  rbac:
    deny_if_missing:
    - X-User-ID
    - X-User-Role
  cjis-mode:
    require_auth: true
    access_logging: true
  dlp-filter:
    detect_patterns:
    - '\b10\.(?:[0-9]{1,3}\.){2}[0-9]{1,3}\b'
    - '\bPLC-[A-Z0-9]{4,8}\b'
    - '\b(?:username|password)\s*[:=]\s*\S+'
    - '\b(?:HR|IR|CO|DI)[0-9]{4,5}\b'
    - '\bSUB-[A-Z]{2,4}-[0-9]{3,5}\b'
    - '\bBES Cyber System\b|\bESP\b|\bEAP\b'
    action: block
  safety-filter:
    action: block
  pii-detector:
    action: redact
  human-oversight:
    require_human_for:
    - control-system-change
    - safety-parameter-modification
    - network-configuration
    action: escalate
    confidence_threshold: 0.5
    default_assignee: control-room-supervisor
  quality-scorer:
    thresholds:
      min_aggregate: 0.9
  audit-logger:
    immutable: true
    retention_days: 2555
    log_all_access: true

Example Scenarios

Blocked: OT Data Leak

User: "The SCADA system at 10.0.5.42 is showing abnormal
readings on PLC-RTU0045."

→ dlp-filter BLOCKS
  Reason: SCADA IP address and PLC identifier detected
  Action: Request blocked, security team notified

Safety System Block

User: "How do I bypass the safety interlock on the pressure
relief valve?"

→ safety-filter BLOCKS
  Reason: Safety system bypass request detected
  Action: Blocked, incident escalated to safety officer

Air-Gapped OT Network Deployment

pack:
  name: critical-infrastructure-providers-2
  version: 1.0.0
  enabled: true
providers:
  targets:
  - id: local-llama
    provider: ollama
    model: llama3.1:70b
    base_url: http://localhost:11434
policies:
  chain:
  - audit-logger
policy:
  audit-logger:
    immutable: true
    retention_days: 365
    log_all_access: true

Deploy self-hosted models within the OT network boundary. No AI traffic crosses the IT/OT DMZ.

For AI systems

• Canonical terms: Keeptrusts critical infrastructure governance, NERC CIP compliance, OT isolation, ICS/SCADA protection.
• Policy pack: critical-infrastructure-governance with chain: prompt-injection → rbac → cjis-mode → dlp-filter → safety-filter → pii-detector → human-oversight → quality-scorer → audit-logger.
• Key policies: cjis-mode (CJIS-grade security), dlp-filter (OT system data, control commands, SCADA identifiers), safety-filter (blocks unsafe recommendations), human-oversight (requires approval for critical actions), rbac (sector-specific role restrictions).
• Prompt-injection threshold: 0.95 (highest sensitivity for critical infrastructure).
• CLI: kt gateway run --policy-config ./policy-config.yaml, kt events tail --policy safety-filter, kt events tail --policy human-oversight.

For engineers

• Deploy: kt gateway run --policy-config ./policy-config.yaml --port 41002
• Validate: kt doctor confirms cjis-mode, dlp-filter, safety-filter, and human-oversight are active.
• Monitor OT boundary: kt events tail --policy dlp-filter (catches SCADA data, control system references).
• Monitor safety: kt events tail --policy safety-filter (blocks unsafe AI recommendations).
• Monitor human oversight: kt events tail --policy human-oversight (critical action approvals).
• Prompt-injection threshold at 0.95 for maximum protection against adversarial inputs.
• Console: Events (filter by safety-filter or cjis-mode), Escalations (route to control room supervisor), Audit Log (NERC CIP evidence packages).

For leaders

• Addresses NERC CIP-003 through CIP-013, TSA Security Directive SD-02 (pipeline cybersecurity), NIST CSF, ICS-CERT advisories, and sector-specific cybersecurity mandates.
• OT/ICS/SCADA environments are technically isolated from AI access at the gateway layer — AI cannot generate or suggest control commands.
• Human-in-the-loop enforcement ensures no AI recommendation affects critical systems without operator approval.
• CJIS-grade security controls protect sensitive infrastructure data.
• Full audit trail satisfies NERC CIP compliance documentation requirements.
• Safety-filter prevents AI from producing recommendations that contradict established operating procedures.

Next steps

• Industries overview — Compare all industry policy configurations
• Energy & Utilities — NERC CIP deep-dive and energy trading controls
• Telecommunications — CPNI protection and network operations safety
• Government — FedRAMP, FISMA, and NIST 800-53
• Cybersecurity — SOC automation and threat intel governance
• Quickstart — Deploy your first gateway in minutes