Answer box
If you're a CISO at a hospital, a Medicare Advantage payer, a digital-health vendor, or a pharma company deploying AI in clinical or operational workflows, you sit at the intersection of six regulatory frameworks: CMS-4201-F and the Feb 2024 FAQ on AI in coverage decisions, CMS-0057-F prior auth APIs, the WISeR Innovation Center model, HTI-1's predictive DSI transparency requirements, Section 1557's anti-discrimination obligations, and HIPAA. Each tells you something different, and together they form a stack no other vertical faces in this combination. This post is the CISO-facing introduction to the stack; the gated 25-page playbook walks through every operational control your team needs to demonstrate and the artifacts your auditor will ask for.
Why healthcare AI security is its own category
In every other regulated industry, AI security is a derivative of broader compliance frameworks — you take NIST AI RMF or ISO 42001 and apply it to your environment. Healthcare is different. Healthcare gets six regulatory bodies pointing at AI specifically, each with its own enforceable rule.
Most CISOs in healthcare have not yet built a unified program for this stack. They handle HIPAA (the security team owns it), CMS rules (compliance owns them), HTI-1 (the EHR vendor handles it), Section 1557 (legal handles it). The result is a coverage map with seams — and the seams are where AI incidents live.
The playbook unifies all six into one operational program. Built around the 20-control library from our unified compliance crosswalk, it shows healthcare CISOs which controls cover which obligations.
The six pieces of the healthcare AI stack
1. CMS-4201-F + the February 2024 FAQ
The codified rule restricting Medicare Advantage organizations' use of AI in coverage determinations. Three operative constraints: algorithmic coverage decisions must be patient-specific; algorithms cannot terminate care on their own; algorithms cannot shift coverage criteria by ingestion. Full coverage in our CMS AI Guidance briefing.
2. CMS-0057-F (Interoperability and Prior Auth)
The FHIR API rule reshaping prior auth data flows. Decision SLAs take effect January 1, 2026; full APIs January 1, 2027; first public performance reporting March 31, 2026. AI tools touching prior auth operate against published SLAs and public metrics — the most regulated AI surface in healthcare.
3. The WISeR Innovation Center Model
Six-year CMMI pilot, January 2026 - December 2031, six states. First major test of AI-assisted prior auth in traditional fee-for-service Medicare. Human-in-the-loop required by design. WISeR's evolution will set precedent for AI in federal healthcare programs.
4. HTI-1 (ASTP/ONC)
The Decision Support Interventions certification criterion. Predictive AI features in certified EHRs (Epic, Oracle Cerner, athenahealth, MEDITECH) must publish 31 source attributes covering training data, validation, bias mitigation. Effective January 1, 2025. If you deploy AI inside a certified EHR, this disclosure pack flows through to your procurement obligations.
5. Section 1557 (HHS OCR)
The May 2024 final rule with patient-care decision-support tool provisions effective May 1, 2025. Makes covered entities liable for discriminatory outputs of third-party AI tools. CISOs cannot disclaim discrimination through procurement language alone.
6. HIPAA
Always present. Training data, inference logs, audit trails, and memory stores are all PHI subject to HIPAA's privacy and security rules.
What the playbook covers (25 pages)
The gated PDF walks through:
- The 20-control library mapped to each of the six healthcare-specific obligations above
- Six common AI use cases in healthcare (clinical decision support, prior auth automation, claims processing, fraud detection, patient communication, drug discovery) and which obligations each triggers
- The HTI-1 disclosure pack — what to demand from your EHR vendor, what your auditor will want to see, sample questions
- The Section 1557 risk assessment — how to score third-party AI tools for discrimination risk, the documentation that satisfies "reasonable efforts to identify and mitigate"
- CMS audit preparation — the 12 artifacts that auditors are asking for and how to prepare them
- Incident response considerations — what changes when the AI is operating under HIPAA's breach notification rules and Section 1557's discrimination obligations
- The 90-day implementation plan — staged rollout from inventory to compliant production
Who the playbook is for
- CISOs at hospitals and health systems building AI security programs
- Compliance officers at Medicare Advantage payers preparing for OIG audits
- Heads of GRC at digital-health vendors answering procurement questionnaires from health-system customers
- Pharma security and compliance teams deploying AI in clinical trials and commercial operations
- Vendor risk teams evaluating AI tools for healthcare environments
If your organization touches Medicare, Medicaid, or a CMS Innovation Center model — and uses AI anywhere in coverage, payment, prior authorization, clinical decisions, or care delivery — the playbook applies.
Why we built this as a vertical playbook
Generic AI security guidance is good for top-of-funnel education. Vertical playbooks are what CISOs actually share with their audit committees. A hospital CISO needs a document that names CMS-4201-F by number, references Section 1557 obligations specifically, and treats HIPAA as the baseline assumption — not a generic "AI security for enterprises" guide that the audit team will dismiss as not applicable.
Each vertical we publish (this is the first of five — financial services, defense, pharma, and tech/SaaS to follow) takes the same approach: ground the playbook in the actual rule numbers and audit artifacts of the regulated industry, written for the CISO who has to defend the program to people who don't have time for theory.
Download the playbook
Download the Healthcare AI Security Playbook (free, 25 pages) — 5-field form, instant PDF delivery.
FAQ
What healthcare entities does the playbook cover?
Hospitals, health systems, Medicare Advantage organizations, Medicaid managed care plans, digital-health vendors, pharma and life-sciences companies, and any covered entity under HIPAA deploying AI in healthcare workflows.
Does the playbook address FDA SaMD requirements?
Yes, in the chapter on AI tools that meet the device definition. The playbook explains where CMS coverage rules require (not replace) FDA clearance and how the two overlap operationally.
Is the playbook free?
Yes. Five-field form (name, work email, company, role, organization type) in exchange for the PDF.
Will there be playbooks for other industries?
Yes. Healthcare is the first; financial services, pharma, defense, and tech/SaaS will follow on a roughly quarterly cadence through Q4 2026. Subscribers to the playbook list will be notified when new ones publish.
Can my compliance team use this in audit prep?
Yes. The playbook is structured around the 12 artifacts auditors are asking for. It's auditor-presentable as-is.
Does the playbook reference specific vendors?
No. The playbook focuses on regulatory obligations and operational controls, not vendor evaluations. If you want vendor-specific guidance, contact our team or use the Enterprise Agent RFP framework.
Related reading: CMS AI Guidance for Healthcare: The 2026 CISO Compliance Briefing · One Map to Rule Them All: A Unified Crosswalk Between NIST AI RMF, ISO 42001, and the EU AI Act · Pillar Hub: AI Compliance Evidence.