Healthcare organizations are adopting AI faster than their security and governance programs can keep up.
More than 70% of healthcare organizations surveyed in the 2026 Healthcare AI Readiness Index have already begun using AI, yet fewer than 40% have detailed policies governing employee AI use. Only 42% of payers and 32% of providers say they are very prepared to respond to AI-assisted cyberattacks. Meanwhile, 60% of payers and 64% of providers report employees using unauthorized or non-integrated “shadow AI” tools.
That creates a difficult question for healthcare CISOs, CIOs, compliance leaders, and AI teams:
How do you prove that the AI your organization develops, deploys, or relies on is actually secure?
Having HIPAA controls in place is important. Having an AI governance policy is important. Having an ISO 42001-certified management system can strengthen organizational governance.
But none of these, on their own, necessarily prove that an AI system has been tested against AI-specific cybersecurity threats.
That is where HITRUST AI Security becomes increasingly relevant.
Healthcare AI Is Moving Faster Than Security
AI is already being integrated into clinical workflows, patient engagement, diagnostics, administrative operations, revenue cycle management, cybersecurity, and healthcare analytics. That expansion is also creating new attack surfaces. A healthcare AI system may interact with:
- Protected health information (PHI)
- Electronic health records
- Clinical applications
- APIs
- Cloud platforms
- RAG databases and knowledge repositories
- Third-party AI models
- AI agents and plugins
- Identity and access systems
- Patient-facing applications
- Internal business workflows
The security implications are significant.
Proofpoint’s October 2026 healthcare research found that 60% of surveyed organizations do not have controls governing sensitive data used in AI systems, while only 42% use security measures specifically designed to protect AI systems, agents, and generative AI. Fewer than one-third have full visibility into employees’ use of unsanctioned AI tools.
At the same time, 45% of healthcare organizations have already embedded AI into cybersecurity, patient care, or both, while another 30% plan to adopt it.
The result is a widening gap between AI adoption and AI security assurance.
And healthcare cannot afford to treat that gap as an ordinary technology risk. A compromised AI system could expose sensitive patient information, produce manipulated outputs, make unauthorized decisions, or provide attackers with a pathway into connected healthcare systems.
What Is HITRUST AI Security?
HITRUST AI Security Assessment and Certification is designed to provide validated cybersecurity assurance for deployed AI systems.
Unlike a general AI governance framework, HITRUST AI Security uses prescriptive, AI-specific security controls and independent validation to assess whether organizations are addressing cybersecurity threats associated with AI. The assessment can apply to different types of AI, including:
- Generative AI
- Predictive AI and machine learning
- Rule-based AI
It considers AI-specific components such as the deployed model, AI platform, specialized AI infrastructure, datasets, agents, plugins, RAG implementations, and supporting technology layers. HITRUST’s AI Security requirements address threats including:
- Prompt injection
- Data poisoning
- Model poisoning
- Model extraction and theft
- Model inversion
- Compromised third-party models or datasets
- Supply chain attacks
- Excessive agency
- Sensitive information disclosure
- Harmful code generation
- AI-specific threat management
- AI security testing and red teaming
HITRUST’s current specification provides for up to 44 AI security-specific requirements, depending on how the assessment is tailored. For healthcare organizations, that distinction matters. The question isn’t simply whether the organization has an AI policy.
It is whether the AI system itself can withstand realistic cybersecurity threats.
HITRUST AI Security vs. ISO 42001: What Is the Difference?
The easiest way to understand the distinction is:
HITRUST AI Security = AI cybersecurity assurance
ISO/IEC 42001 = AI management and governance
ISO/IEC 42001 establishes an Artificial Intelligence Management System, or AIMS. It provides an organizational structure for managing AI through policies, accountability, risk management, lifecycle processes, monitoring, and continual improvement.
HITRUST AI Security takes a much more focused approach to the cybersecurity of AI systems.
HITRUST itself describes its AI Security Certification as complementary to ISO/IEC 42001 rather than a replacement for it.
That distinction is important because healthcare organizations often need both organizational governance and technical security assurance.
What Does ISO 42001 Give a Healthcare Organization?
ISO 42001 is useful when the primary question is:
“How does our organization govern AI responsibly and consistently?”
An ISO 42001 program can help establish:
- AI governance structure
- Roles and responsibilities
- AI policies
- AI risk management
- AI objectives
- AI lifecycle controls
- Data governance
- Monitoring and performance evaluation
- Documentation
- Continual improvement
- Accountability for AI-related decisions
For a healthcare organization with multiple AI use cases, vendors, departments, and stakeholders, this organizational structure is valuable. It creates a repeatable way to identify, manage, and monitor AI-related risks rather than addressing every AI deployment as a separate project.
But ISO 42001 is intentionally broad.
It is not primarily a technical security testing standard. An organization could have an established AIMS and still need to test whether its AI chatbot is vulnerable to prompt injection, whether a RAG application exposes PHI, or whether an AI agent has excessive permissions.
What Does HITRUST AI Security Add?
HITRUST AI Security addresses the question:
“Can we demonstrate that the AI system is protected against cybersecurity threats?”
That can include evaluating:
AI threat management
Organizations need to understand what can go wrong within the AI architecture and how those threats are being mitigated. HITRUST specifically incorporates threat modeling and security evaluations such as AI red teaming.
Prompt injection
Can malicious instructions manipulate the AI into ignoring its intended behavior, revealing sensitive information, or taking unauthorized actions?
Data and model poisoning
Could malicious or manipulated data affect the AI system’s behavior?
Model extraction and inversion
Can an attacker recover sensitive information about the model or infer information from it?
Excessive agency
Can an AI agent perform actions beyond what it should be allowed to perform?
This is increasingly important as healthcare organizations move from AI that simply generates information toward AI that can interact with applications and execute workflows.
Sensitive information disclosure
Can the AI reveal PHI, proprietary information, credentials, internal instructions, or other sensitive data through crafted interactions?
Supply chain security
What happens if a third-party model, dataset, plugin, agent, software component, or AI platform introduces risk into the healthcare environment?
These are not theoretical governance questions. They are technical attack paths.
Why Healthcare Needs More Than HIPAA for AI
HIPAA remains foundational for protecting PHI and regulating how covered entities and business associates handle protected health information. But AI introduces security questions that are not solved simply by having a HIPAA compliance program. Consider a healthcare organization deploying an AI-powered clinical assistant. The organization may have:
- Access controls
- Encryption
- Audit logging
- Privacy policies
- Business associate agreements
- Workforce training
- Incident response procedures
Yet the AI system could still be vulnerable to:
Prompt injection → unauthorized retrieval → PHI exposure
Or:
Compromised document → RAG retrieval → malicious instruction → unauthorized AI action
Or:
Excessive agent permissions → manipulated AI workflow → unauthorized system change
Traditional compliance controls remain necessary. But AI-specific security testing is needed to understand how the AI layer changes the attack surface. That is why HITRUST’s AI Security approach adds AI-specific requirements to the broader cybersecurity environment rather than treating the AI system as an isolated component.
Does HITRUST AI Security Replace ISO 42001?
No.
And healthcare organizations should be cautious about treating these as competing certifications.
ISO 42001 and HITRUST AI Security solve different problems.
ISO 42001 establishes a management system for AI.
HITRUST AI Security provides security-focused assurance for AI systems.
Think of it this way:
ISO 42001 asks:
How does the organization govern AI?
HITRUST AI Security asks:
How well is the AI system protected against cybersecurity threats?
NIST AI RMF asks:
How does the organization identify, assess, and manage AI risks?
These approaches can work together.
In fact, HITRUST explicitly positions its AI Security Certification as complementary to ISO/IEC 42001.
When Should a Healthcare Organization Consider HITRUST AI Security?
HITRUST AI Security becomes particularly relevant when AI is becoming part of a business-critical or regulated healthcare workflow. Consider it if your organization:
- Develops AI-powered healthcare products
- Provides AI-enabled healthcare services
- Operates an AI chatbot that handles sensitive information
- Uses AI with PHI
- Deploys clinical AI
- Uses RAG against internal healthcare data
- Deploys AI agents with access to enterprise systems
- Provides AI platforms to healthcare customers
- Needs independent AI security validation
- Faces customer requirements for AI security assurance
- Already maintains a HITRUST e1, i1, or r2 program
- Needs to demonstrate stronger AI security assurance to healthcare partners
HITRUST states that its AI Security Certification is designed primarily for providers of AI systems, including AI application and platform providers. It can be added to HITRUST e1, i1, or r2 assessments.
That makes it particularly relevant for healthtech companies and healthcare technology providers that need to demonstrate that their AI systems are secure to enterprise healthcare customers.
What If You Already Have HITRUST e1, i1, or r2?
You do not necessarily need to start from scratch. HITRUST designed its AI Security requirements to build on an underlying HITRUST CSF assessment. The AI Security Certification can be incorporated with an e1, i1, or r2 assessment, with up to 44 additional AI security-specific requirements depending on scope.
That creates an important opportunity for healthcare organizations that already have a mature HITRUST program. Instead of creating an entirely separate AI security framework, organizations can extend their existing security assurance model to address the risks introduced by AI. The underlying principle is simple:
AI-specific threats are additive to traditional cybersecurity threats.
Your existing security controls still matter. Your AI environment needs additional controls on top of them.
What Should a Healthcare AI Security Assessment Actually Test?
Before pursuing certification or making a major AI deployment, organizations should understand the complete AI attack surface.
- AI application security: Test the application surrounding the model, including authentication, authorization, APIs, session management, input handling, and application logic.
- Model security: Evaluate model-specific risks such as extraction, inversion, manipulation, unsafe behavior, and model misuse.
- Prompt injection: Test direct and indirect prompt injection through user inputs, uploaded documents, RAG content, emails, web content, and other untrusted sources.
- RAG security: Evaluate retrieval controls, document permissions, vector databases, data segregation, and the possibility of malicious content influencing model behavior.
- AI agent security: Determine what an agent can access, which tools it can invoke, what permissions its identity has, and whether human approval is required for high-impact actions.
- Sensitive data protection: Test whether PHI, PII, credentials, proprietary information, system instructions, or other restricted data can be extracted through AI interactions.
- AI supply chain: Assess third-party models, APIs, datasets, plugins, agents, platforms, and other dependencies.
- Logging and monitoring: Determine whether the organization can detect suspicious prompts, abnormal model behavior, unauthorized tool calls, sensitive-data extraction, and other AI-specific events.
- Threat modeling: Map realistic attack paths across the AI architecture rather than testing individual components in isolation.
- Remediation and validation: A security assessment should not end with a vulnerability report. Critical findings should be remediated and retested to establish whether the controls actually work.
The Healthcare AI Security Gap Is Already Visible
The latest healthcare data makes the urgency difficult to ignore. Healthcare organizations are adopting AI rapidly, but many are still building the controls needed to secure it. The 2026 Healthcare AI Readiness Index found that:
- More than 70% of healthcare organizations have started using AI.
- Fewer than 40% have detailed policies governing employee AI use.
- Only 42% of payers and 32% of providers feel very prepared for AI-assisted cyberattacks.
- 60% of payers and 64% of providers report unauthorized or non-integrated shadow AI use.
- 90% expect their AI and cybersecurity investment to increase over the next year.
The latest Proofpoint/Ponemon healthcare research adds another layer to the problem. 97% of surveyed U.S. healthcare organizations experienced at least one cyberattack in the previous 12 months, and 74% of those affected reported disruption to patient care.
Now combine that existing threat environment with AI systems that can access sensitive healthcare data and increasingly interact with other applications. The question becomes less about whether healthcare AI will be targeted and more about whether organizations can prove they are prepared when it is.
HITRUST AI Security Is Not Just Another Certification
For healthcare organizations, assurance increasingly has to mean more than having another compliance badge. Customers, partners, boards, regulators, and security teams want evidence that AI systems are actually being protected.
That is where independent validation becomes valuable.
HITRUST AI Security combines prescriptive AI-specific controls with independent assessment and certification. Its requirements are also harmonized with sources including NIST, ISO, and OWASP. For healthcare organizations already operating within the HITRUST ecosystem, this can provide a natural extension of an established security and compliance program. For healthtech companies selling AI-powered products, it can provide a way to demonstrate AI security assurance to customers that increasingly want more than a generic security questionnaire.
Where Does Accorian Fit?
AI security cannot be separated from healthcare cybersecurity, compliance, and assurance. Accorian brings these capabilities together across HITRUST, healthcare cybersecurity, AI security, penetration testing, AI risk, AI governance, and compliance. Its AI security capabilities include:
- AI Security Assessments
- AI Chatbot Penetration Testing
- LLM Security Testing
- Prompt Injection Testing
- AI Red Teaming
- Agentic AI Security Assessments
- AI Threat Modeling
- Third-Party AI Security Validation
- AI Risk Assessments
- AI Security Governance
- HITRUST for AI Systems
- NIST AI RMF
- ISO 42001
Accorian’s approach is particularly relevant for healthcare because the assessment needs to look beyond the model itself and understand how AI interacts with patient data, healthcare applications, APIs, identities, knowledge repositories, and downstream systems.
Accorian also provides HITRUST CSF assessment services across e1, i1, and r2, supported by GORICO, its AI-enabled GRC platform with direct HITRUST MyCSF integration.
That creates a more connected path from:
AI discovery → AI risk assessment → AI security testing → HITRUST readiness → evidence management → validated assurance
rather than treating each requirement as a separate compliance exercise.
Should You Pursue HITRUST AI Security, ISO 42001, or Both?
There is no universal answer.
If your primary objective is to establish organization-wide AI governance, ISO 42001 may be the more relevant starting point.
If your primary objective is to demonstrate validated cybersecurity assurance for an AI system, HITRUST AI Security may be the stronger fit.
If your organization develops or deploys high-impact AI in healthcare, the strongest strategy may be to use both.
ISO 42001 can establish how AI is governed.
HITRUST AI Security can demonstrate how AI security is validated.
And broader risk-management approaches such as NIST AI RMF and ISO/IEC 23894 can help address risks beyond cybersecurity. The goal is not to collect certifications.
The goal is to build an AI program that can withstand scrutiny.
Is Your Healthcare AI Ready for Security Assurance?
Healthcare AI is moving into increasingly sensitive workflows. If your AI can access PHI, interact with clinical systems, retrieve internal information, make recommendations, call APIs, or take autonomous actions, traditional compliance alone may not provide enough assurance. You need to know:
- What can go wrong?
- Can an attacker exploit it?
- What could they access or control?
- Can your organization detect the attack?
- And can you prove that your controls work?
HITRUST AI Security provides one path toward answering those questions through AI-specific controls, independent validation, and certification. For healthcare organizations building serious AI programs, the question is no longer simply:
“Are we using AI responsibly?”
It is:
“Can we prove that the AI we depend on is secure?”
Build AI Security That Can Withstand Scrutiny
Accorian can help healthcare organizations and healthtech companies assess AI-specific threats, strengthen security controls, prepare for HITRUST AI Security, and connect AI security with broader governance and compliance programs.
Don’t wait for a customer, auditor, or attacker to expose the gap in your AI security program.
Talk to Accorian about your HITRUST AI Security readiness.
