Peer-reviewed · 2026
Runtime Authority-Aware Threat Modeling
A seven-dimensional framework covering intent, context, authority, tool capability, memory, delegation, and action composition.
Read overview ↗← Home
AI security research
Research & frameworks
Practical research on agent authority, threat modeling, runtime controls, and enterprise AI security.
Current work
Map authority.
Control action.
Autonomous systems need controls that remain effective as agents access tools, data, identities, memory, and delegated capabilities.
Technical report · 2026
AI Runtime Threat Matrix
A taxonomy and control framework for risks across agent workflows, including the RAAI runtime control loop.
Read overview ↗Runtime model
RAAI Runtime Security Model™
Continuous assessment, authorization, action, and audit for workflows whose authority changes over time.
Threat modeling
Agentic AI Runtime Attack Tree
Paths through prompts, memory, tools, APIs, permissions, identities, and downstream actions.
Enterprise practice
ANNAM Framework
Architecture, runtime protection, monitoring, and governance controls for enterprise AI systems.