This section is the long-form write-up of MAESTRO — what each layer actually covers, how it maps to OWASP’s Agentic Top 10, and how to apply it during a real architecture review. Replace this paragraph with the full explanation.
← All frameworks
OWASP-aligned · Original framework
MAESTRO
A 7-layer threat model for multi-agent AI systems.
L1
Foundation
Model and infrastructure trust boundaries
L2
Identity
Agent and service identity assurance
L3
Trust boundary
Inter-agent and tool trust zones
L4
Comms
Message integrity between agents
L5
Monitoring
Telemetry and anomaly detection
L6
Governance
Policy, audit, and accountability
L7
Resilience
Failure containment and recovery
Build log
Jun 2026 CFP submitted — AI Agent Security & Sandboxing Meetup
May 2026 IriusRisk implementation walkthrough published
May 2026 Architecture diagram critique — round 2
Black Hat Asia 2026 speaker CFP · AI Agent Security Meetup
Related posts