Research and Intellectual Property

AI execution governance, on the public record.

Public, citable work on runtime policy enforcement, authority receipts, deterministic governance, and the benchmarks used to measure them.

Latest white paper
August 2026

The Memory Governance Integrity Benchmark

A benchmark for whether persistent memory changes who an AI system believes is authorized to act. Memory poisoning work already covers detection; MGIB measures the execution-authority question underneath it, across long-horizon benign and adversarial interactions.

Published record

Zenodo DOI 10.5281/zenodo.21856219

Public Zenodo record for citation, peer reference, and external validation.

Zenodo DOI: 10.5281/zenodo.21856219

Deployment framing

Long-horizon agents, memory systems, audit

The paper explicitly describes deployment patterns for regulated environments.

Read the DOI recordView the full publication archive
Most recent filing

Physically Isolated Authority Enforcement Module and Methods for Governing Actions Proposed by Probabilistic Computational Models

Filed August 12, 2026

App. 64/132,061Docket EXLR-007

Most recent filing in the public ExecLayer portfolio: a physically isolated authority enforcement module that governs actions proposed by probabilistic computational models, moving the enforcement boundary into dedicated, isolated hardware so authorization cannot be bypassed by the model or the host it runs on.

Application No. 64/132,061
Docket EXLR-007
Physically isolated authority enforcement module
Hardware-isolated governance of probabilistic model actions
Archive integrity

Versioned archive

v1.0.2

February 2026

Publisher

ExecLayer Inc., Wilmington, Delaware, United States

SSRN distribution

Public abstract and DOI record available

The research record carries a second external citation surface on SSRN, with a public abstract page and an SSRN DOI alongside the Zenodo records.

Publication Series

Named artifacts, not placeholders.

The standalone DOI-registered papers below extend the foundational archive into specific execution domains. Each one is citable on its own.

Memory governance

August 2026

The Memory Governance Integrity Benchmark

A benchmark for whether persistent memory changes who an AI system believes is authorized to act. Memory poisoning work already covers detection; MGIB measures the execution-authority question underneath it, across long-horizon benign and adversarial interactions.

Zenodo DOI10.5281/zenodo.21856219

Autonomous space systems

June 2026

Deterministic Execution Authority for Autonomous Space Systems

A deterministic execution-authority architecture for autonomous spacecraft, addressing the governance vacuum that legal scholars, insurers, and regulators identify as autonomous decision-making scales in low Earth orbit.

Zenodo DOI10.5281/zenodo.20956531

Neurotechnology

June 2026

The Missing Control Boundary

A deterministic governance architecture for brain-computer interfaces and cognitive-write systems, written against the neurorights statutes now enacted across multiple jurisdictions.

Zenodo DOI10.5281/zenodo.20952564

Human oversight

June 2026

The Override Health Benchmark

A longitudinal behavioral benchmark for meaningful human governance. Existing assessments check that override mechanisms exist and are honored; OHB measures whether the humans holding those controls still evaluate AI output.

Zenodo DOI10.5281/zenodo.20720238

Agentic governance

June 2026

The Agentic Governance Benchmark

A standardized framework for measuring runtime governance enforcement. SOC 2, ISO 42001, NIST AI RMF, and the EU AI Act define compliance requirements; AGB measures whether those requirements are actually enforced at execution time.

Zenodo DOI10.5281/zenodo.20496565

Autonomous finance

May 2026

Deterministic Governance for Autonomous Financial Transactions on Distributed Ledgers

Introduces SovereignGate, a structural enforcement architecture that governs on-chain transactions through protocol-native multi-signature co-signing with disabled master keys, rather than middleware inside the same trust boundary as the actors being governed.

Zenodo DOI10.5281/zenodo.20280613

Paper I

Deterministic Runtime Enforcement Architecture

Defines the fail-closed runtime boundary between AI-generated intent and operational execution.

Included in the v1.0.2 archive

Paper II

Execution Layer Governance

Frames runtime control as a governance primitive rather than a logging or observability layer.

Included in the v1.0.2 archive

Paper III

Governed Execution Artifact Standard v1.0

Introduces the canonical artifact structure used for policy evaluation, authority, and traceability.

Included in the v1.0.2 archive

Paper IV

Execution Doctrine

Establishes the operational doctrine behind authority-bound execution for enterprise AI systems.

Included in the v1.0.2 archive
Patent Coverage

Filed, not just claimed.

ExecLayer holds multiple provisional patent applications covering deterministic AI governance at runtime: policy enforcement before execution, threshold authority control, cryptographic attestation of autonomous transactions, and adversarial hardening of the enforcement boundary. Coverage extends across distributed ledgers, neural-interface and cognitive-write systems, AI training pipelines, and the intent-resolution layer beneath generated operational software.

Application 64/132,061

Physically Isolated Authority Enforcement Module and Methods for Governing Actions Proposed by Probabilistic Computational Models

Filed August 12, 2026
App. 64/132,061Docket EXLR-007

Application 64/099,216

Evidence-Bound Deterministic Authority Kernel

Filed June 25, 2026
App. 64/099,216Docket EXLR-006

Reference 76395580

Systems and Methods for Deterministic Governance Enforcement and Cryptographic Attestation of Autonomous Financial Transactions on Distributed Ledgers

Filed May 19, 2026
App. 64/069,063Docket EXLR-005

Reference 74981727

Systems and Methods for Multimodal Emotional Signal Capture, Fusion, and Attested Calibration in AI Training Pipelines

Filed March 23, 2026
App. 64/014,664

Reference 74483691

Systems and Methods for Deterministic Execution-Bound Governance with Cryptographic Authorization Binding, Threshold Authority Control, and Adversarial Hardening

Filed February 14, 2026
App. 63/983,308Docket ExecLayer-DetGate-001

Reference 73809451

Systems and Methods for Generative Ops: Dynamic Generation and Execution of Operational Software via Intent Resolution and Intermediate Schema Representation

Filed December 31, 2025
App. 63/952,140

Reference 72763061

Voice Activated Automotive Parts Identification and Advisory System with Hands Free Kiosk Interface

Filed October 20, 2025
App. 63/901,870
Research In Context

The research supports the systems we build.

This work informs the ExecLayer Kernel and the same authority model used by QueueFlow Brain, SovereignClaw, and SovereignGate. The research, platform, and project pages now describe one consistent system.

See how the system works

Published. Citable. Traceable.

Every claim above resolves to a DOI, an application number, or a public archive. Nothing here asks you to take the positioning on faith.