Last week I joined CHAI’s AI Cybersecurity Work Group kickoff as an AI Governance scholar-operator and author of Ungoverned: A Practical Guide to AI Minimum Viable Governance (2026). This was operational reality, not theoretical positioning.
The economics flipped.
According to Anthropic’s AI cyber team: 2.3 attacks per day on healthcare in H1 2026—up 14% from the prior half. +35% targeting healthcare businesses: billing firms, device makers, distributors. Median ransom demand: $310K. Outliers reached nine figures.
Why I’m here.
I focus on AI governance, not technical cybersecurity. I see AI governance as a leadership challenge. Organizational leaders must ensure their technical teams’ expertise translates into practical guidance that non-technical leaders and institutions can adopt and act on.
My interest sits at the intersection of AI cybersecurity and AI governance: how organizations, boards, and executive teams should oversee, manage, and stay accountable for AI-related cyber risk—including frontier-model threats, model risk, data privacy, vendor accountability, and the governance gaps that legacy frameworks leave ungoverned.
When autonomous agents multiply across healthcare infrastructure, the accountability question shifts: who owns the decision to deploy this agent, who verified the risk, and who has authority to stop it if the threat model changes? That’s governance. That’s where I can help.
What this workgroup seeks to create.
This team seeks to produce actionable guidance: agent inventory standards, egress control protocols, 48-hour clinical downtime scenarios, vendor accountability frameworks. These are concrete moves that help close the visibility gap before the window shuts in healthcare.
AI Governance as Leadership can translate that technical rigor into a governance framework that boards and executive teams can operationalize—not as a compliance checkbox, but as a leadership discipline.
https://www.chai.org/cybersecurity-workgroup
Questions for your teams:
Can you inventory every AI agent deployed—internal and vendor—right now? Do you have egress control and logging at the agent boundary? Have you drilled 48-hour clinical downtime scenarios? Have you asked your top vendors the five accountability questions on agent governance? Is AI running against your own code and infrastructure before external threat actors do?
If you cannot answer yes to all five, your institution is not governing the deployment. You’re discovering the gap after.
Grateful to:
@CHAI, led by CEO @Brian Anderson and team including @Anthony DiDonato, for this work. @Anthropic’s cybersecurity experts—@Syed Mohiuddin (Head of Healthcare), @Rob Bair (Cyber and National Security Policy), @Josh Adelsberg (Healthcare GTM)—and health system security leaders: @Alan Berry (Centene), @Ryan Winn (AdventHealth), Isaiah Nathaniel (Delaware Valley). And to every leader at this table who understands: governance gets built into deployment decisions from the start, not retrofitted after.
Disclosure: This analysis draws on the CHAI kickoff and Anthropic’s AI cybersecurity presentation materials. Drafted and refined using AI-assisted tools for synthesis and clarity. Final editorial judgment and responsibility remain with the author. For informational and educational purposes only. Not legal, regulatory, or compliance advice. Non-vendor. Non-partisan. Doctoral rigor, not advocacy.
© 2026 Dr. Freddie Seba. All rights reserved.
#AIGovernance #ResponsibleAI #AIMVG #Ungoverned #Healthcare #BoardGovernance #EnterpriseAI #CIO #AILeadership #AIAccountability #AIRisk #CHAI #InstitutionalLeadership
Leave a Reply