CISOs are being asked to approve agentic AI use cases that didn't exist six months ago—often without governance playbooks, risk models, or peer benchmarks to work from. Jason Clinton, Deputy CISO at Anthropic, breaks down how to evaluate the capability-versus-risk tradeoff at the center of every one of these decisions.
What You'll Learn
Clinton will cover which agentic capabilities are ready for enterprise deployment, which ones aren't, and what controls actually matter when the system you're governing can take actions on its own. Using Claude Cowork as a case study, he'll demonstrate how governance primitives (scoped access, network egress controls, SIEM-routed telemetry, admin-paced rollout) can ship alongside the capabilities themselves. The session goes beyond any single product to provide a working framework for how security leaders should think about agent governance right now.Key Topics:
- A mental model for evaluating agentic use cases—breaking down requests into specific actions, scopes, and blast radii you can reason about
- The governance primitives that matter: scoped permissions, egress control, telemetry routing, admin-paced rollout
- Where CISOs already deploying agentic AI at enterprise scale are prioritizing, and the patterns that separate thoughtful rollouts from the ones generating incidents
Why This Matters for Healthcare
Agent governance is particularly critical in healthcare where autonomous actions can affect patient care workflows, PHI access, and clinical documentation systems. Healthcare organizations evaluating AI assistants for clinical coding, discharge summary generation, or administrative task automation need frameworks for scoping agent permissions, monitoring agent behavior, and understanding blast radius when agents can read from EHRs or write to scheduling systems.Clinton's governance primitives framework—scoped access, egress control, telemetry, staged rollout—maps directly to the controls healthcare security teams need to deploy before approving agentic use cases. This is the conversation healthcare CISOs should be having internally before the next department requests approval to pilot an AI agent.
Event Details
- Date: May 12, 2026, 10:00 AM PT (Recorded - registration provides access to replay)
- Duration: 60 minutes with live Q&A
- Speaker: Jason Clinton, Deputy CISO at Anthropic
- Format: Live webinar with Q&A, recording available to registrants within 48 hours
- Cost: Free
- Registration: Anthropic Webinar Page
Training & Events posts highlight learning opportunities relevant to healthcare AI security practitioners. These are educational resources, not endorsements.