Healthcare organizations are rapidly deploying agentic AI systems to streamline operations and improve patient care. Simultaneously, security researchers are sounding the alarm about a widening gap between innovation and protection.
The autonomous nature of AI agents, which are designed to perform complex tasks with minimal human oversight, is creating a host of new vulnerabilities in an industry already under siege from cyber threats. Case in point: Healthcare was the most-breached industry in 2024.
There’s no denying the power of AI in healthcare. But as agentic systems become more sophisticated and independent, they’re fueling a massive attack surface that traditional security frameworks weren’t designed to address. It’s time for healthcare organizations to take a hard look at their approach to AI security and compliance.
New Security and Compliance Challenges
Amidst this changing landscape, new threats are emerging. One such example is zero-click attacks that target agentic AI systems. Unlike traditional attacks that require user interaction (like clicking a link or opening an attachment), zero-click exploits compromise agents through passive exposure to malicious data or inputs that agents process automatically.
This is particularly concerning in healthcare, where AI agents continuously process data from multiple sources (e.g., lab results, imaging files, insurance claims). For example, an attacker could embed malicious code in a seemingly innocuous document that an AI agent processes as part of its normal workflow. The attack surface for zero-click exploits and the potential for damage is substantial.
But new threats are inevitable. The problem is that most organizations deploying agentic AI aren’t equipped to mitigate them. In addition to strengthening their AI security posture, healthcare organizations must also consider how these systems impact compliance.
Frameworks like HIPAA were designed for a world where data access and processing were controlled by humans. Today, healthcare organizations are grappling with challenges related to accountability, oversight, and how existing regulations apply to autonomous systems.
The only way forward is to adopt a strategy that recognizes AI agents as autonomous risk entities. Here’s what that looks like:
Implement Continuous Behavioral Monitoring
Continuous behavioral monitoring specifically designed for AI agents is critical. This goes beyond simply logging access: organizations need the ability to analyze behavioral patterns, flag deviations from expected operational norms, and detect signs of manipulation or compromise in real-time. Agents should be monitored as distinct entities with their own risk profiles—not just extensions of the users who deploy them.
Establish Clear Trust Boundaries For Agents
Security teams must establish clear boundaries and constraints on agent capabilities through technical controls, not just policies. This includes implementing strict data minimization principles, limiting access to specific datasets and systems, and building kill switches that can immediately revoke agent permissions if suspicious activity is detected.
Develop AI Security Testing Protocols
As a best practice, healthcare organizations should also develop comprehensive testing protocols for AI security. This can include adversarial testing, red-teaming, penetration testing specifically targeting agent manipulation, and regular security audits that assess both technical controls and operational practices around AI deployment.
Determine AI Governance Frameworks
It’s crucial to establish clear governance frameworks that address AI-specific risks. This includes things like defining who is accountable when an agent makes a decision that leads to a security breach, how to conduct investigations when autonomous systems are involved, and what documentation and oversight requirements apply to AI deployments handling protected health information.
Account For Agentic AI Compliance
Security leaders should engage closely with compliance teams to develop interpretations of HIPAA requirements that account for agentic AI. This may require seeking guidance from the Department of Health and Human Services on how existing regulations apply to autonomous systems, or advocating for regulatory updates that provide clearer frameworks for AI in healthcare.
The healthcare industry is at a critical juncture. Agentic AI offers genuine promise for improving care delivery, reducing administrative burden, and addressing workforce challenges. But realizing these benefits safely demands more robust security. Organizations that move forward with agentic AI must do so fully aware of the risks, with appropriate security controls in place, and with an ongoing commitment to continuously evolving their security posture. By implementing the strategy above, healthcare organizations can usher in the next evolution of AI-driven care—safely and securely.
