In AI We Trust? Securing the Future, One Agent at a Time
Panelat Open Source SecurityCon North America 2025 · on Monday, 2025-11-10· at 16:15· Building B | Level 3 | B302-303·
summary
Core cloud‑native controls apply but require extension: examples include using workload identity and JWT delegation (an RFC exists), enforcing zero trust is table stakes with cryptographic attestations, and reducing LLM calls to avoid exponential risk. Practical tools discussed are Solo.io’s KAgent/Agent Gateway, Presidio for prompt/PII checks, OPA‑style policies at MCP gateways, and human approvals for write actions. Many MCP implementations lack scoped tokens, so default secure defaults in open source agent platforms matter.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 3:53AI vs microservice security
- 10:44Zero trust for AI agents
- 18:50Agent permission delegation
- 26:58AI agent security standards
- 31:33Judgment in multi-agent systems
- 35:45Policy engines for MCP
- 39:39Closing & Q&A
related talks
42mZero Trust for Autonomous Agents: Isolating AI Workloads on Kubernetes
01h 11mKeynote - Securing AI: Building Trust in the Era of AI
23mZero-Trust Architecture for Agentic AI: Securing Multi-User Access and Third-Party Integrations
21mSecurity in the Age of AI & Copilot
17mDon't give your agent the master key
24mAI Governance for developers: A practical framework for deploying autonomous coding agents safely…
27mFrom Perimeter to Silicon: Why Agentic AI Demands a New Security Model
06mAI Agents, Sovereignty & the Next Decade of Innovation