Theoretical analysis demonstrates a five-layer architecture for warrant containment in legal AI systems, highlighting limits of automated dependency tracking.
Legal artificial intelligence is commonly evaluated at the level of the answer, but consequential legal systems require a different object of governance: the continuing warrant for institutional reliance. A legal conclusion may have been well grounded when produced and yet become unfit for reliance after a statute is amended, a precedent is displaced, an authorization expires, a jurisdiction changes, or an evidentiary premise is withdrawn. Placing Vitaly Reznik's Zero-Trust Logic and warrant-containment experiments in dialogue with CriptoIUS, this paper proposes a five-layer architecture for living normative memory: authority admission, dependency representation, formal containment, runtime admissibility, and institutional reliance. Formal dependency containment can compute what follows from a recorded change, but legal and institutional governance must decide whether the change occurred, what authority it carries, and whether the result may legitimately support action. Technical and interpretive boundary. The formal results discussed in the paper are machine-checked classifications over recorded dependencies. They do not interpret legal norms, establish the validity of legal authority, detect hidden dependencies, prove genuine independence between external grounds, or replace the institutional judgment required for reliance. Public architecture boundary. CriptoIUS is used as a conceptual research frame. The paper does not disclose commercial workflows, internal schemas, scoring rules, thresholds, routing logic, or private product architecture.
No takes yet. Share an insight, caveat, or question.
Ignacio Adrián LERER (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: