Systems architecture evaluation demonstrates a portable evidence framework for autonomous AI agents, indicating robust cryptographic auditability across platforms.
AEGIS defines a portable evidence interface that carries AI-agent execution records from declared cross-platform capture through sequence commitment, producer-independent verification, independently governed custody, auditor-controlled procedure and report, and forum judgment. ARC1-ARC6 define the architecture; authenticated data structures, signatures, encodings, key discovery, and anchoring remain replaceable profile choices. AEGIS v1 instantiates ARC2-ARC3 with AEGIS-FRAME-v1, SHA-256, Ed25519, JCS, a signed certificate, and the V1-V4 acceptance rule. The current version makes the independent local audit-repository profile and structured verification receipt explicit. It records subsequent convergence toward receipts, witnesses, custody, audit stores, and external-auditor roles. It also documents attributed downstream engineering reuse in IBM Client Engineering's public DFAH-Bench repository, whose hash-chain and certificate modules identify AEGIS as their source and are connected to a replay-log bundle builder and tests. Version note. Version 4.0 aligns the public record with the manuscript submitted to Computer Law & Security Review on 26 August 2026. Stable artifact citations use concept DOIs: the main article remains at 10.5281/zenodo.18955102; Supplemental Material A is at 10.5281/zenodo.21348560; and Supplemental Material B is at 10.5281/zenodo.21348562. Earlier versions remain permanently available in their version chains.
No takes yet. Share an insight, caveat, or question.
Alex Chengyu Li (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: