Skip to content
shadowiq
Regulation · HITRUST CSF + AI

Healthcare AI with PHI redaction inline — and signed proof.

HITRUST is the healthcare control language. ShadowIQ extends it to generative AI with inline PHI redaction, signed evidence, and controls that pass HITRUST r2 validation.

What this is

Summary

HITRUST CSF with AI combines the HITRUST Common Security Framework with AI-specific controls for healthcare environments, including inline PHI (Protected Health Information) redaction at the AI gateway, cryptographic evidence aligned to HIPAA, 42 CFR Part 2, and NIST 800-53.

How it fits · explainer

The crosswalk: article → control → signed evidence.

HITRUST · ARTICLEREQUIREMENTSHADOWIQ CONTROLSIGNED EVIDENCE01.yTeleworkingZero-trust AI gatewaySigned access records09.aaAudit loggingEvidence ledgerSigned append-only log11.aReporting info security eventsOCSF incident exportSigned incident recordsAI-1AI inventoryRegistry + discoverySigned AI BOMAI-3AI risk assessmentEval engine + DPIASigned eval + DPIACROSSWALK · 5 SHOWN · FULL MAP IN /DOCS/COMPLIANCE · SIGNED ED25519
Where it hurts

You've heard this one before.

  • PHI leaking into AI assistants that were never in HITRUST scope.
  • r2 validation missing AI-specific evidence.
  • HIPAA BAA scope ambiguity for LLM providers.
  • Business Associate audit requests with no ready artifacts.
What we do about it

Three moves.

  1. 1
    Inline PHI redaction.

    18 HIPAA identifiers + custom schemas detected with context. Redact, tokenize, or deny before the model sees them.

  2. 2
    HITRUST control mapping.

    AI-specific evidence mapped to HITRUST v11 controls. Accept / compensating / inherit statuses with signed artifacts.

  3. 3
    BAA-ready.

    Signed records of subprocessor data flows, provider BAAs, and model-residency policies. Your BA audit package prefills itself.

Outcomes

Numbers, not adjectives.

0
PHI egress events · 2.1M monthly calls
r2
validation-ready
18
HIPAA identifiers detected
Full crosswalk

HITRUST article → ShadowIQ control → signed evidence.

Article
Requirement
ShadowIQ control
Signed evidence
01.y
Teleworking
Zero-trust AI gateway
Signed access records
09.aa
Audit logging
Evidence ledger
Signed append-only log
11.a
Reporting info security events
OCSF incident export
Signed incident records
AI-1
AI inventory
Registry + discovery
Signed AI BOM
AI-3
AI risk assessment
Eval engine + DPIA
Signed eval + DPIA
Frequently asked

Asked, answered, sourced.

Most major providers do under specific plans: OpenAI Enterprise, Anthropic Enterprise, Azure OpenAI, AWS Bedrock, Google Vertex AI (enterprise tier). ShadowIQ signs a BAA in Growth+ plans and enforces BAA-scope routing at the gateway.

Named-entity recognition tuned on healthcare data plus regex for structured identifiers (MRN, account numbers). Customer-schema detectors add organization-specific identifiers.

ShadowIQ operates under SOC 2 Type II + ISO 27001; HITRUST r2 validation is in progress for Q3 2026. We produce BA-ready attestations in the interim.

Ready to see the signet in motion?

Your 30-minute demo. A signed audit trail by the end of it.

We'll wire ShadowIQ into one live workload, block a prompt injection in real time, and hand you a cryptographic receipt — before the meeting ends.