BLOG

by Morgan Hague

The integration of Artificial Intelligence (AI) and Generative AI (GenAI) is rapidly shifting the CISO's role from guardian of data to Architect of Digital Trust. For healthcare organizations, this shift is critical. AI is both a force multiplier for efficiency and a complex new attack surface that introduces unique risks extending beyond traditional cybersecurity into patient safety and clinical equity.

To manage this, CISOs’ most critical strategic action is not to build a new program from scratch but leverage their existing program into managing evolving AI risk in an efficient way. One accessible and widely defensible method outlined below is to integrate the NIST AI Risk Management Framework (AI RMF) into the existing governance structures of HIPAA and the NIST Cybersecurity Framework (CSF).

This outline highlights the primary concerns CISOs face today with the advent of AI sprawl, and easily accessible solutions for adapting existing programs or compliance mechanisms.

The Three Unique Risks: Beyond the Perimeter

AI risks in a healthcare setting are fundamentally different from traditional IT risks, often tied directly to clinical outcomes and legal liability. CISOs must prioritize these three categories:

• Algorithmic Bias (The Equity Risk): AI models, trained on historically incomplete or non-representative health data, can perpetuate systemic inequities. This translates to a risk of disparate health outcomes (e.g., misdiagnosis in underrepresented populations). This is not just an ethical issue; it is a patient safety issue that exposes the organization to significant legal liability under anti-discrimination laws.
• Data Hallucination & Clinical Error (The Accuracy Risk): AI systems, especially Large Language Models (LLMs), can generate outputs that are factually incorrect, misleading, or entirely fabricated (known as “hallucination”). If this error infiltrates clinical documentation, diagnostic support, or treatment plans, the result is a direct and immediate threat to patient care.
• Privacy by Inference (The HIPAA Strain): AI’s core utility is to infer non-obvious facts. This power strains the HIPAA Privacy Rule’s “minimum necessary” standard. AI can potentially predict highly sensitive private information (PHI) even from seemingly benign or de-identified datasets (e.g., inferring a condition from mouse movements or search patterns).

The Strategic Blueprint: Operationalizing the NIST AI RMF

The NIST AI Risk Management Framework (AI RMF) is the prevailing standard for responsible AI governance. The RMF is intentionally designed for integration into existing enterprise risk programs like the NIST CSF.

CISOs can leverage the AI RMF’s four core functions to map out their governance strategy.

Actionable Integration: Mapping AI Risk to Existing Frameworks

Rather than completely rebuilding your compliance program; you need to augment it. The NIST AI RMF or a designated and suitable alternative is key to aligning AI initiatives with the regulatory demands you already face, particularly in healthcare.

Integrating with HIPAA and Risk Analysis

The HIPAA Security Rule mandates a thorough Risk Analysis to identify threats and vulnerabilities to ePHI. Consider these key areas:

• Augmentation: Use the Map and Measure functions of the AI RMF to conduct a rigorous, defensible AI Risk Analysis that satisfies OCR expectations and provides meaningful insights to internal stakeholders. This analysis must specifically identify the risk of privacy by inference and the potential for algorithmic bias that can lead to security events or inaccurate patient care.
• Incident Response: Update Incident Response Plans to include scenarios for an AI-induced patient harm event or a compromise from an adversarial attack designed to manipulate an AI model.

Augmenting the NIST Cybersecurity Framework 2.0

The NIST CSF provides an accessible structure to operationalize AI governance within your technical controls – including the following:

• Govern (G.AP): The Govern function of the AI RMF becomes your direct implementation of the CSF's Governance category, establishing your organization's risk acceptance criteria for AI.
• Identify (ID.SC): The Map function (AI inventory) provides the context necessary for the Supply Chain Risk Management category (ID.SC).
• Protect (PR.DS): The Measure function's focus on data quality and security directly informs the Data Security category (PR.DS) by applying controls to the massive, sensitive training datasets AI requires.

The CISO’s New Frontier: Vendor Risk Management

Most AI in healthcare is consumed as a third-party service. Your current Vendor Risk Management (VRM) program must evolve to address this:

• New Due Diligence: Add AI-specific requirements to your assessment questionnaires. Ask vendors about their model governance, the diversity of their training data, and their procedures for detecting and mitigating hallucinations and bias.
• Contractual Mandates: Ensure contracts with AI vendors include clauses for auditability and shared responsibility for model performance, transparency, and the secure handling of inferred or generated PHI.

Conclusion

The CISO who masters AI risk management will not only secure their organization but will enable faster, safer, and more ethical innovation, becoming a true strategic partner to the clinical mission.

Meditology, as a healthcare-exclusive cybersecurity and compliance firm, can help integrate AI risk management into your organization’s existing cybersecurity program by leveraging a specialized suite of services that map directly to the NIST AI Risk Management Framework (AI RMF) and HIPAA requirements.

Our support is structured around governance, program strategy, technical validation, and third-party risk management.

Furthermore, Meditology’s risk philosophy distinguishes between specific Third-Party Risk (like a vendor mishandling patient records) and the broader, systemic Supply Chain Risk (like a compromised piece of hardware from a manufacturer). This granular understanding is critical for CISOs adopting AI, as they must manage both the security of the specific AI vendor and the upstream components (e.g., open-source models, training data, cloud infrastructure) that constitute the AI's supply chain.

About the Author

Morgan is an experienced security and emerging technologies consultant, with varied expertise across information security, organizational governance, and IT audit practices. As the leader of the Strategic Risk Consulting and AI/ML service lines at Meditology, he has led and contributed to hundreds of consulting engagements across public and private entities.

Since 2019, he has served as lead architect and product owner of an innovative risk quantification, analysis, and reporting solution utilizing MITRE ATT&CK and similar authoritative sources to establish a data-driven and dynamic mechanism to assess, report on, and manage organizational risk – supporting a variety of premier healthcare organizations, including the nation’s largest hospital system.

Morgan is currently an executive board member with InfraGard Atlanta, an effort lead with the OWASP AI Exchange, and serves as an external advisor for AI and automation working groups at some of the nation’s premiere providers.