Research Brief · Governing Relationship State
A record can stay true as history while becoming an inadequate basis for what happens next
Micheal Charles Preble · Operant Dyad — Paper 4
The Question
What legal and institutional framework should govern the continued use, transfer, restriction, revalidation, retirement, and reconstitution of persistent relationship state once its legal significance can change on a different timeline than the record itself?
The Problem
Persistent AI systems can accumulate corrections, preferences, contextual interpretations, prior decisions, restrictions, and authority representations that shape later work. That state can remain available even after the factual, legal, or institutional basis for relying on it has changed. An employee may have authority to approve a procedure while occupying one role and later move to another; the record of their approval can remain completely true even after they no longer possess authority to approve today.
The Contribution
Argues the relevant legal object is not ownership of AI memory but lawful continuing reliance on persistent state. Proposes a purpose-limited governance interest in “Covered Relationship State” — state materially used to characterize, coordinate with, attribute authority to, or make consequential decisions about a person. Within that category, “Authority-Bearing State” is state relied on as evidence of permission, consent, delegation, or role; a “Material Predicate” is any fact, authority, role, purpose, or consent element whose validity matters to whether “Consequential Reliance” is presently justified; and the actor with sufficient control to determine, authorize, continue, suspend, or materially alter that reliance holds “Legally Relevant Control Over Reliance.”
How the Argument Works
A comparative U.S./EU legal synthesis, not an empirical study. Draws on GDPR Articles 18–20 (which already separate storage from use, and portability from a general property right), the FCRA's reinsertion regime, federal records-preservation law, HIPAA authorization rules, the EU Data Act's switching provisions, and real case law (Van Patten v. Vertical Fitness, Reyes v. Lincoln Automotive, Herbert Construction v. Continental Insurance) used as structural analogues, not as AI-specific precedent. The central move is separating historical validity (was the representation once true or authorized?) from current operative validity (does it remain an adequate basis for reliance now?).
Why It Matters
The paper's governing principle: a durable record can preserve what was once true without determining what remains authorized now. Forgetting everything on demand destroys accountability and institutional memory; treating everything remembered as perpetually authoritative converts persistence into unaccountable control. The proposed framework aims at governable persistence — not maximal retention, and not maximal deletion.
Evidence Status
Evidence status: Conceptual legal and institutional analysis; published preprint. A model governance architecture whose adoption, territorial scope, and enforcement would depend on jurisdiction-specific legal vehicles — not a claim of a harmonized transnational rule.
What This Does Not Claim
- Does not require constant consent or recurring confirmation of benign persistent state.
- Does not require universal deletion, or make every derivative traceable by legal fiction.
- Does not override institutional records, third-party rights, security, confidentiality, trade secrets, public duties, or independent legal authority.
- Does not propose automatic damages for every procedural defect.
- Does not claim a single institutional vehicle (federal statute, model legislation, sectoral regulation) has already been proven superior.
Open Questions
“Institutional overlap” — whether a broad framework would conflict with mature sector-specific systems, or whether a model law implemented through existing regulators is more proportionate — is described as “the most important unresolved policy question.” Whether the framework should ultimately be generalized across sectors depends on empirical prevalence, existing remedies, and implementation cost, which the paper explicitly leaves open.
Read / Cite the Research
Read the canonical paper record → · Full paper PDF · View on SSRN
Preble, Micheal Charles. “Governing Relationship State: A Legal Framework for Human-Controlled AI Continuity” Available at SSRN 7396659, 2026.