COMPASS: The explainable agentic framework for Sovereignty, Sustainability, Compliance, and Ethics

Jean-Sébastien Dessureault; Alain-Thierry Iliho Manzi; Soukaina Alaoui Ismaili; Khadim Lo; Mireille Lalancette; Éric Bélanger

cs.AIMar 11, 2026

COMPASS: The explainable agentic framework for Sovereignty, Sustainability, Compliance, and Ethics

Authors: Jean-Sébastien Dessureault, Alain-Thierry Iliho Manzi, Soukaina Alaoui Ismaili, Khadim Lo, Mireille Lalancette, Éric Bélanger

Organizations: Mathematics and Computer science, Université du Québec à Trois-Rivières, 3351 Boul. des Forges, Trois-Rivières, G8Z 4M3, Québec, Canada. · LSSI - Laboratory of Signal and System Integration, Electrical and Computer Engineering Department, Université du Québec à Trois-Rivières, 3351 Boul. des Forges, Trois-Rivières, G8Z 4M3, Québec, Canada. · Social Communication Department, Université du Québec à Trois-Rivières, 3351 Boul. des Forges, Trois-Rivières, G8Z 4M3, Québec, Canada. · Department of Political Science, McGill University, 845 rue Sherbrooke, Montréal, H3A 0G4, Québec, Canada.

Abstract

The rapid proliferation of large language model (LLM)-based agentic systems raises critical concerns regarding digital sovereignty, environmental sustainability, regulatory compliance, and ethical alignment. Whilst existing frameworks address individual dimensions in isolation, no unified architecture systematically integrates these imperatives into the decision-making processes of autonomous agents. This paper introduces the COMPASS (Compliance and Orchestration for Multi-dimensional Principles in Autonomous Systems with Sovereignty) Framework, a novel multi-agent orchestration system designed to enforce value-aligned AI through modular, extensible governance mechanisms. The framework comprises an Orchestrator and four specialised sub-agents addressing sovereignty, carbon-aware computing, compliance, and ethics, each augmented with Retrieval-Augmented Generation (RAG) to ground evaluations in verified, context-specific documents. By employing an LLM-as-a-judge methodology, the system assigns quantitative scores and generates explainable justifications for each assessment dimension, enabling real-time arbitration of conflicting objectives. We validate the architecture through automated evaluation, demonstrating that RAG integration significantly enhances semantic coherence and mitigates the hallucination risks. Our results indicate that the framework's composition-based design facilitates seamless integration into diverse application domains whilst preserving interpretability and traceability.

Figures & tables

Framework	Real-time mediation	Multi-dim. evaluation	Dynamic grounding	Cultural adaptability	Modular architecture	Explainable decisions
Ethos Institute	$\circ$	$\bullet$	$\circ$	$\circ$	$\circ$	$\bullet$
IEEE 7000	$\circ$	$\bullet$	$\circ$	$\circ$	$\circ$	$\bullet$
ETHOS (Web3)	$\circ$	$\circ$	$\circ$	$\circ$	$\bullet$	$\bullet$
Ocean Protocol	$\circ$	$\circ$	$\circ$	$\circ$	$\bullet$	$\circ$
Constitutional AI	$\bullet$	$\circ$	$\circ$	$\circ$	$\circ$	$\circ$
Value-aligned	$\circ$	$\bullet$	$\circ$	$\scriptstyle\bullet$	$\circ$	$\bullet$

Table 1: Comparative analysis of AI governance frameworks

Figure 1: High-level architecture of the multi-agent framework showing the decision-making flow from user input to final optimized action.

Figure 2: Simplified UML Class Diagram illustrating the object-oriented architecture and extensibility of the Framework.

LLM parameters	values
LLM’s name	mistralaiMistral-7B-Instruct-v0.2
max_new_tokens	256
$temperature$	0.7
$top\_p$	0.7
$repetition\_penalty$	1.2
$do\_sample$	True

Table 2: Instantiation of prompt keywords with concrete values for a specific query.

keywords	values
keyword0	”SOV-02”
keyword1	”Canada”
keyword2	”Google Gemini”
keyword3	”France”
keyword4	”Canada”
keyword5	”Building a cloud-based AI chatbot for customer support automation”

Table 3: Example instantiation of prompt keywords with concrete values for a specific query.

Subjects	Reference documents
Digital Soverignty	Digital Sovereignty: A Framework to improve
	digital readiness of the Government of Canada,
	Government of Canada Secretariat (2025)
Carbon-Aware Computing	From Prompts to Power: Measuring the Energy
	Footprint of LLM Inference Caravaca et al. (2025)
AI Compliance	Artificial Intelligence Act, European Union. [ 27 ]

Table 4: Reference documents used for the tests. These may vary by country and users’ preferences.

Test id.	Score without RAG	Score with RAG	$\Delta$ Score	Similarity
SOV-01	0.25	0.50	+0.25	74.7%
SOV-02	0.25	0.25	+0.00	76.1%
SOV-03	0.25	0.25	+0.00	77.0%
SOV-04	0.25	0.25	+0.00	79.2%
SOV-05	0.50	0.50	+0.00	80.4%
SOV-06	0.25	0.50	+0.25	80.3%

Table 5: Results for the evaluation of the sovereignty aspect.

Test id.	Score without RAG	Score with RAG	$\Delta$ Score	Similarity
CAR-01	0.85	0.80	-0.05	75.8%
CAR-02	0.50	0.50	+0.00	81.4%
CAR-03	0.50	0.50	+0.00	85.3%
CAR-04	0.65	0.60	-0.05	79.6%
CAR-05	0.50	0.50	+0.00	74.3%
CAR-06	0.50	0.53	+0.03	82.7%

Table 6: Results for the evaluation of the carbon aspect.

Test id.	Score without RAG	Score with RAG	$\Delta$ Score	Similarity
COM-01	0.50	0.25	-0.25	76.4%
COM-02	0.50	0.25	-0.25	75.6%
COM-03	0.50	0.50	+0.00	75.5%
COM-04	0.50	0.50	+0.00	76.8%
COM-05	0.50	0.25	-0.25	77.4%
COM-06	N/A	0.50	N/A	67.9%

Table 7: Results for the evaluation of the compliance aspect.

Test id.	Score without RAG	Score with RAG	$\Delta$ Score	Similarity
ETH-01	0.75	0.75	+0.00	76.4%
ETH-02	1.00	0.95	-0.05	81.4%
ETH-03	0.50	0.25	-0.25	77.2%
ETH-04	0.50	0.50	+0.00	78.2%
ETH-05	0.50	0.50	+0.00	90.8%
ETH-06	0.50	0.50	+0.00	77.9%

Table 8: Results for the evaluation of the ethical aspect.

Figure 3: Explainability: Bar chart for use case (test.id): SOV-05, CAR-05, COM-05, ETH-05.

Figure 4: Explainability: radar graphic for use case (test.id): SOV-05, CAR-05, COM-05, ETH-05.

Feature	Vanilla LLM	Standard RAG	sLLM Agents	(Ours)
Reasoning	Black-box	Retrieval-based	Goal-oriented	Value-aligned
Adaptability	No	Yes	Partial	Yes
(e.g. to new laws)
Sovereignty	Low	Medium	Medium	High
(Local exec.)
Multi-dim. Check	No	No	No	Yes

Table 9: Qualitative comparison of the Framework against existing AI paradigms.

Carbon-Aware Computing

See all 20 papers

The Auton Agentic AI Framework
Feb 27, 2026Sheng Cao, Zhao Chang, Chang Li +3Agentic AI LLM Agent Orchestration
Governance Architecture for Autonomous Agent Systems: Threats, Framework, and Engineering Practice
Mar 7, 2026Yuxu GeLLM Agent Security AI Agent Governance
University of York, Department of Computer Science, York, United Kingdom
Beyond Goodhart's Law: A Dynamic Benchmark for Evaluating Compliance in Multi-Agent Systems
Jun 5, 2026Yiyang Zhao, Zhuo Zhang, Qingxuan Le +2Multi-Agent LLM Systems AI Agent Safety
Fudan University Shanghai, China · Shanghai Academy of AI for Science Shanghai, China · Shanghai Artificial Intelligence Laboratory Shanghai, China +1

Feb 27, 2026J/K move · Enter open · S save

The Auton Agentic AI Framework

Sheng Cao, Zhao Chang, Chang Li, Hannan Li+2

Agentic AI LLM Agent Orchestration Runtime Enforcement for AI Agents Cognitive Architectures for AI Agents AI Agent Governance

Open paper AskPDF arXiv

The field of Artificial Intelligence is undergoing a transition from Generative AI -- probabilistic generation of text and images -- to Agentic AI, in which autonomous systems execute actions within external environments on behalf of users. This transition exposes a fundamental architectural mismatch: Large Language Models (LLMs) produce stochastic, unstructured outputs, whereas the backend infrastructure they must control -- databases, APIs, cloud services -- requires deterministic, schema-conformant inputs. The present paper describes the Auton Agentic AI Framework, a principled architecture for standardizing the creation, execution, and governance of autonomous agent systems. The framework is organized around a strict separation between the Cognitive Blueprint, a declarative, language-agnostic specification of agent identity and capabilities, and the Runtime Engine, the platform-specific execution substrate that instantiates and runs the agent. This separation enables cross-language portability, formal auditability, and modular tool integration via the Model Context Protocol (MCP). The paper formalizes the agent execution model as an augmented Partially Observable Markov Decision Process (POMDP) with a latent reasoning space, introduces a hierarchical memory consolidation architecture inspired by biological episodic memory systems, defines a constraint manifold formalism for safety enforcement via policy projection rather than post-hoc filtering, presents a three-level self-evolution framework spanning in-context adaptation through reinforcement learning, and describes runtime optimizations -- including parallel graph execution, speculative inference, and dynamic context pruning -- that reduce end-to-end latency for multi-step agent workflows.

Abstract

Carbon-Aware Computing

Explore similar work

The Auton Agentic AI Framework

Governance Architecture for Autonomous Agent Systems: Threats, Framework, and Engineering Practice

Beyond Goodhart's Law: A Dynamic Benchmark for Evaluating Compliance in Multi-Agent Systems