Enterprises are adopting AI agents faster than security teams can find them, making agentic AI security an urgent CISO priority in 2026. Agents hold credentials, call tools, and act autonomously, so one exploited agent operates with its user's full access at machine speed. In episode 926 of CXOTalk, Anand Oswal, EVP of AI and Network Security at Palo Alto Networks, describes a practical sequence: discover every agent first, then apply identity, runtime, and governance controls through a centralized AI gateway.
What the conversation covers
- The four surfaces where agents run: enterprise, SaaS platforms, endpoints, and the browser
- Why vibe coding agents and third-party MCP servers expand the supply chain attack surface
- How browser agents holding live sessions and cookies enable identity impersonation
- New runtime threats: prompt injection, memory poisoning, tool misuse, and model denial of service
- Agent identity: deciding what an agent can connect to, with what permissions, for how long
- Why a centralized AI gateway and a unified platform beat stitched-together point products
Episode Participants
Anand Oswal is the Executive Vice President and General Manager of AI and Network Security at cyber security leader Palo Alto Networks.
Michael Krigsman is a globally recognized analyst, strategic advisor, and industry commentator known for his deep expertise in business transformation, innovation, and AI leadership.
0 Commentaires