PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
June 2, 20260 citationsOpen Access

Tacked Up: Harnessing Agentic AI

View Full Paper
SFS. Michelle Farr

Key Points

  • The aim is to address the security challenges of autonomous AI by focusing on the control layer surrounding the agents.
  • Developed a threat model for agentic AI systems.
  • Outlined three planes of agentic zero trust and assurance conditions.
  • Investigated how mediation fails under increased load.
  • Proposed a reference architecture for evaluating mediation in autonomous agents.
  • Identified critical failure points in agentic zero trust frameworks.
  • Demonstrated that existing security principles still apply to modern AI systems.

Abstract

Autonomous AI agents are reaching production faster than the controls meant to govern them. This paper relocates the agentic security problem from the model to the control layer that surrounds it, the trust and security harness, and shows that this layer is a present-day instance of a specification more than fifty years old: the reference monitor of Anderson (1972) and the protection principles of Saltzer and Schroeder (1975). It supplies a threat model, three planes of agentic zero trust, a set of testable assurance conditions, and a capacity-aware account of how mediation quietly fails under load. The aim is a reference architecture for judging whether an autonomous agent is actually mediated before it can act on enterprise systems.

Ask AI
Helpful
Bookmark
Share
View Full Paper

Cite This Study

S. Michelle Farr (2026) studied this question.

synapsesocial.com/papers/6a1e732830b38c64201b666fhttps://doi.org/10.5281/zenodo.20471174
Ask AI
Helpful
Bookmark
Share
View Full Paper