Research & Publications

Stephen Joseph Voss — Qensai

U.S. Provisional Patent Applications 63/985,184 (Feb 18, 2026) and 63/991,184 (Feb 26, 2026) — 98 claims

The Structural Gaps

The increasing deployment of autonomous and multi-agent computational systems has exposed structural gaps in existing governance models. Current approaches rely on access control, policy enforcement, and post-hoc auditing — mechanisms that presuppose correct behavior rather than structurally ensuring it.

Authorization-Appropriateness Gap

Existing models determine whether a principal is authorized but not whether the authorization remains appropriate given current system state.

Governance Externality

Governance is applied to systems rather than embedded in system identity. It remains an external constraint, not an intrinsic property.

Temporal Proxy

Time is used as a proxy for state validity. A credential valid when issued may be inappropriate when used if governance context has changed.

Fixed Vocabulary

Governance frameworks enforce only what their designers anticipated. Unanticipated operational phenomena escape governance entirely.

Evidence-Reality Gap

Audit records document what was reported, not what physically occurred. Evidence is detached from the physical reality it describes.

Self-Governance Absence

The governance apparatus does not measure or constrain its own institutional health. It governs everything except itself.

Passive Security

Hardware security relies on static manufacturing properties rather than the dynamic operational state of the machine.

Coordination-Governance Conflation

Multi-agent governance is treated as a coordination problem rather than a structural integrity problem.

What the Framework Does

The constitutional governance architecture addresses these gaps by embedding governance into the identity, state, and interaction model of every participating component. Incompatible states are structurally unreachable rather than policy-blocked. The contribution is architectural — describing what the framework does and the structural mechanisms by which it achieves its properties.

Hash Identity Structure

Every assembly carries a multi-tier hash encoding identity, state, configuration, and composition. Interaction between assemblies is governed by compatibility of corresponding tiers — a geometric property of the hash space, not a permission granted by an authority.

Temporal Entropy Authorization

Credentials are derived from the temporal entropy of the physical execution environment at the moment of generation. Entropy sources are consumed in measurement — they do not persist, cannot be queried, and cannot be reconstructed. Every credential is a temporal event, not a stored artifact.

Dual-Key Authority

Two independent authority channels encode fundamentally different governance dimensions — identity and state. No single authority can unilaterally authorize execution. The impossibility of substitution is architectural, not cryptographic.

Computed State Correspondence

Credentials embed a governance state snapshot at production time. At presentation, the system evaluates whether embedded values still correspond to current computed state. There is no revocation list, no expiration timestamp, no external authority.

Expression Layer

Assembly identity is computed on demand from governance tables, sensor readings, and temporal entropy. It exists only during active computation. When computation stops, the expression ceases to exist.

Constitutional Resonance Scoring

A continuous measurement of the governance layer's own institutional health. When health is critically degraded, a structural prohibition on credential production activates — the governance instrument declining to produce output it cannot certify as constitutionally valid.

Integrity Architecture

Hash witness verification, canonical-first reconciliation, split-knowledge state governance, and three-ledger hallucination detection. Agreement across independently maintained records is required for confirmed execution.

Autonomous Vocabulary Extension

The system detects operational anomalies, creates new registry entries when standard terms are absent, and assigns calculated weights based on operational evidence. The governance vocabulary grows through operation under constitutional authority.

Evidence-Based Validation

All execution events are recorded with causal anchoring to consumed temporal entropy. Evidence is structurally bound to the physical moment of creation. Retroactive fabrication requires reconstructing entropy consumed in measurement — a structural impossibility.

Application Domains

Distributed AI governance — multi-assembly systems with heterogeneous authority levels

Multi-agent orchestration — preventing locally authorized but globally incompatible actions

Autonomous cybersecurity — governance at machine speed with constitutional accountability

Integrity-critical enterprise — governance of how AI-generated outputs were produced

AI-native infrastructure — environments designed with governance as a foundational property

Regulated AI deployment — verifiable governance records causally tied to physical reality

Implications for AI Governance Policy

Auditability

Evidence-based validation with causal anchoring provides governance records that are independently verifiable and physically bound to the moments they describe.

Accountability

Multi-tier identity and three-ledger validation produce a complete, tamper-evident record of what each assembly did, under what governance conditions, and with what authorization.

Proportionality

The lane architecture enables governance proportional to operational risk. Low-impact exploration is not burdened with high-impact execution governance.

Self-Limitation

Governance systems can be designed to structurally recognize and enforce their own limitations — a property relevant to autonomous AI systems operating within defined constitutional bounds.

Situated Within

The framework is situated within and departs from established research across multiple domains:

Access Control & Authorization — Sandhu et al. (1996), Ferraiolo et al. (2001), Hu et al. (2014), OAuth 2.0 (Hardt, 2012)

Multi-Agent Systems — Wooldridge (2009), Shoham & Leyton-Brown (2009), Park et al. (2023)

Cryptographic Authorization — Diffie & Hellman (1976), Shamir (1979), PFS (Günther, 1990)

Formal Verification — Clarke et al. (2018), Cousot & Cousot (1977), Harrison (2009)

AI Safety — Amodei et al. (2016), Russell (2019), Bai et al. (2022), Christiano et al. (2017)

Hardware Security & PUFs — Pappu et al. (2002), Herder et al. (2014), Zero-Trust (Rose et al., 2020)

Full technical disclosure including mathematical foundation available under NDA.

Request NDA & Technical Disclosure