Threat models and security architecture for a system built around large numbers of autonomous AI agents with access to internal tools, data, and MCP-style servers.
Guardrails, sandboxing, and permissioning for how agents talk to systems and each other, so they can act without excessive standing privileges.
