Healthcare security teams now face adversaries who use AI to find vulnerabilities, scale social engineering, and move from discovery to exploit faster than ever. With average healthcare ransomware demands approaching $17 million, slow detection and response is a risk healthcare organizations can’t afford.
In this on-demand virtual session, Zach Moreno, AVP Consulting at Bishop Fox, moderates a panel with Farzan Karimi, Deputy CISO at Moderna; Rose Lee, VP and CISO at BlueCross BlueShield of Tennessee; and Matthew B. Welling, Partner at Holland & Knight. The panelists explain how they are preparing healthcare security programs for AI-enabled threats while helping their organizations adopt AI safely.
The conversation covers continuous penetration testing and attack surface management, agentic incident response, safe security testing for medical devices and GxP-regulated manufacturing, AI governance and approved-use policies, data classification, vendor risk, and the legal and regulatory realities of responding to incidents at machine speed. Watch on demand to learn which security fundamentals matter most now, how to protect patient care and operations during an incident, and what healthcare CISOs should prioritize over the next three to five years.
Key Takeaways
- AI changes the speed of attacks, not the fundamentals of defense. Identity, least privilege, data protection, monitoring, and incident response still form the foundation. Nonhuman identities are now part of that foundation, and healthcare organizations should extend their existing cybersecurity programs to cover AI rather than bolting on a separate AI security program.
- Assume compromise and limit the blast radius. A single compromised identity or device shouldn’t take down the organization. Defense in depth, conditional access, and kill switches that let incident response teams terminate any session or account help contain AI-enabled attacks.
- Test medical devices and manufacturing systems without risking patient safety. Emulated environments built from extracted device firmware let teams find vulnerabilities without hitting live systems. Incident response playbooks should also separate corporate systems, where isolation is safe, from GxP-regulated environments, where a human makes the containment decision.
- Use AI to scale security across the enterprise. Moderna’s threat modeling agent conducted security reviews of 687 applications in a single month. That scale depended on groundwork such as a CMDB with clear system ownership, so findings could be routed to accountable teams.
- Set guardrails for AI adoption instead of roadblocks. Define approved AI tools and use cases, limit unnecessary options, and train employees early. Sandboxes with synthetic data give teams room to experiment and reduce shadow AI and the risk of regulated data entering unapproved tools.
- Know where sensitive healthcare data flows. A data classification scheme helps AI agents handle restricted and regulated data correctly. Review default-on AI features in incumbent tools before data leaves your environment unexpectedly.
- Document, follow, and exercise your incident response plan. Regulators review incidents with hindsight, so follow your plan and document any deviation. Tabletop exercises with executives and technical teams, such as a service account compromise in manufacturing or an MFA push-fatigue attack, build muscle memory and expose gaps.
- Plan for strain across the healthcare ecosystem. Forensics firms, outside counsel, and cyber insurance carriers are facing capacity limits. Establish those relationships before an incident, and plan for staff burnout and alternative business processes if critical systems go offline.
- Pilot AI security tools before you buy them. Ask what a tool replaces and what it lets you do that you couldn’t before. If a vendor’s pitch still makes sense with “AI” swapped for an older buzzword, the value may be marketing rather than outcomes.