CONTROLLING AUTONOMOUS AI AGENTS : A Comparative Survey of Runtime Enforcement, Agent Architecture, Multi-Agent Monitoring and Collective Behaviour Detection

The rapid development of autonomous artificial intelligence agents raises a fundamental architectural question: How can human authority be preserved when AI systems increasingly plan, select, execute and adapt their own operational activities? Research addressing this challenge is developing across several distinct areas: runtime authorization, least-privilege access control, execution isolation, formal policy verification, system observability, anomaly detection and multi-agent coordination monitoring. This survey examines representative initiatives in these areas, with particular attention to thedistinction between controlling individual agent actions and identifying collective behaviour emerging through interactions among multiple agents. Nine principal research initiatives are examined: Faramesh, Tether, AgentGuardian,MCP-Secure, IsolateGPT, ACE, SentinelAgent, LumiMAS and Parity Swarm. A separate published systematic review of 58 primary studies is also considered asbackground evidence concerning the distribution of runtime-enforcement research. The survey then examines the ST-T1024 A1–A16 architecture and its complementary A17proposal for detecting emergent collective behaviour. The objective is not to establish the superiority of one architecture. It is to identify existingcontributions, examine their respective boundaries, expose unresolved research questions and establish a basis for independent technical comparison and experimentation. Keywords: Agentic AI; Autonomous AI Agents; Runtime Enforcement; AI Security;Multi-Agent Systems; Collective Behaviour; Swarm Detection; Runtime Monitoring;Deterministic Architecture; Spiking Neural Networks; ST-T1024; A17.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-21
DOI
https://doi.org/10.5281/zenodo.24411043
Primary Topic
Access Control and Trust
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

CONTROLLING AUTONOMOUS AI AGENTS : A Comparative Survey of Runtime Enforcement, Agent Architecture, Multi-Agent Monitoring and Collective Behaviour Detection

M. TAALABI, Team LLM-DIPLOMAT
Zenodo (CERN European Organization for Nuclear Research)
Access Control and Trust
article

CONTROLLING AUTONOMOUS AI AGENTS : A Comparative Survey of Runtime Enforcement, Agent Architecture, Multi-Agent Monitoring and Collective Behaviour Detection

M. TAALABI, Team LLM-DIPLOMAT
article en

Abstract

The rapid development of autonomous artificial intelligence agents raises a fundamental architectural question: How can human authority be preserved when AI systems increasingly plan, select, execute and adapt their own operational activities? Research addressing this challenge is developing across several distinct areas: runtime authorization, least-privilege access control, execution isolation, formal policy verification, system observability, anomaly detection and multi-agent coordination monitoring. This survey examines representative initiatives in these areas, with particular attention to thedistinction between controlling individual agent actions and identifying collective behaviour emerging through interactions among multiple agents. Nine principal research initiatives are examined: Faramesh, Tether, AgentGuardian,MCP-Secure, IsolateGPT, ACE, SentinelAgent, LumiMAS and Parity Swarm. A separate published systematic review of 58 primary studies is also considered asbackground evidence concerning the distribution of runtime-enforcement research. The survey then examines the ST-T1024 A1–A16 architecture and its complementary A17proposal for detecting emergent collective behaviour. The objective is not to establish the superiority of one architecture. It is to identify existingcontributions, examine their respective boundaries, expose unresolved research questions and establish a basis for independent technical comparison and experimentation. Keywords: Agentic AI; Autonomous AI Agents; Runtime Enforcement; AI Security;Multi-Agent Systems; Collective Behaviour; Swarm Detection; Runtime Monitoring;Deterministic Architecture; Spiking Neural Networks; ST-T1024; A17.

Zenodo (CERN European Organization for Nuclear Research)
Access Control and Trust
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon
CONTROLLING AUTONOMOUS AI AGENTS : A Comparative Survey of Runtime Enforcement, Agent Architecture, Multi-Agent Monitoring and Collective Behaviour Detection — M. TAALABI, Team LLM-DIPLOMAT · Zenodo (CERN European Organization for Nuclear Research) (2026) | TGRS Research Map | TGRS