SGAEIA: Secure Governed Autonomous Edge Intelligence Architecture

SGAEIA — Secure Governed Autonomous Edge Intelligence Architecture SGAEIA is an open, vendor-neutral research reference architecture for secure, governed, auditable, and trustworthy autonomous AI systems operating across distributed edge-cloud environments. The project investigates architectural foundations for autonomous and multi-agent AI systems in which technical capability is explicitly distinguished from legitimate authority to act. SGAEIA explores architectural and security principles including bounded and revocable authority, authenticated and constrained delegation, non-human workload identity, Zero Trust runtime enforcement, risk-adaptive authorization, continuous Governance, Risk and Compliance (GRC), Evidence-as-Code, formalized security invariants, distributed trust boundaries, and security-preserving technology substitution. The architecture provides a research framework for reasoning about how autonomous AI agents can operate under explicit governance constraints while preserving traceability, auditability, accountability, least privilege, revocation, and human authority. Research Artifact Version 0.3.4 is a public research artifact and demonstrative implementation of the SGAEIA architecture. The artifact includes specifications, reference code, tests, formal models, adapter profiles, and supporting documentation intended to support architectural analysis, experimentation, reproducibility, and further research. Research Scope SGAEIA addresses research areas including: Agentic AI Multi-Agent Systems Edge AI Distributed Autonomous Systems AI Security Zero Trust Runtime Governance Bounded and Revocable Authority Authenticated Delegation Continuous GRC Evidence-as-Code Security-by-Design Spec-Driven Development Formal Assurance Security-Preserving Technology Substitution Research Status SGAEIA is a research reference architecture and demonstrative implementation. It does not constitute a production-ready framework, production certification, safety certification, regulatory approval, or independent security certification. The TLA+ and Alloy artifacts are formal specifications provided for model-checking and formal-analysis purposes. External model-checker execution is claimed only when separately documented by corresponding evidence. Author Aridio SilvaIndependent Researcher — BrazilORCID: https://orcid.org/0009-0008-2411-6995 Canonical Repository SGAEIA GitHub Repositoryhttps://github.com/aridiosilva/SGAEIA Persistent Identifiers Version 0.3.4 DOIhttps://doi.org/10.5281/zenodo.22557796 All-versions DOIhttps://doi.org/10.5281/zenodo.22557795 License Apache License 2.0