Research
Published research, patent portfolio, and ongoing governance failure analysis. SovereignClaw is built on peer-reviewed foundations.
This page is the evidence layer behind the product. SovereignClaw is positioned as an AI agent safety platform, but the underlying claims are anchored in published research, a patent portfolio, and an ongoing public record of governance failures. If you are evaluating the product technically, pair this page with the architecture and security pages. If you are evaluating it operationally, compare the evidence model to the compliance mappings.
Published Research
The SovereignClaw architecture, security properties, and enforcement model are documented in a research paper available on SSRN and DOI-registered on Zenodo:
What the paper covers
The paper explains why agent safety breaks down when execution is treated as a follow-on concern instead of a runtime boundary. It formalizes deterministic execution control, describes the receipt and ledger model, and outlines the security properties that make unsafe operations structurally unreachable instead of merely discouraged.
Patent Portfolio
Six provisional patent applications are pending with the USPTO, covering deterministic governance enforcement, authenticated intent resolution, and cryptographic attestation of high-risk autonomous actions.
Why the patents matter for buyers
The patent portfolio is not just defensive IP. It also signals where the company believes the core novelty lives: authenticated intent resolution, deterministic governance at runtime, and hardened rails for high-risk enterprise actions. That matters when security and legal teams are assessing whether the platform has a differentiated approach to AI execution control.
Governance Failure Radar
Tracking Every Public AI Governance Failure
ExecLayer documents every public AI governance failure, analyzes root cause, and shows how deterministic runtime controls would have prevented it. That ongoing failure analysis gives teams a practical bridge between theory, architecture, and deployment risk.
Ongoing Analysis
Follow ongoing governance failure analysis and SovereignClaw development on the ExecLayer Substack:
How to use this material in evaluation
The fastest way to use this section in a buying process is to read the paper for the core model, review the radar for concrete failure cases, and then connect both to the deployment tiers and evaluation process. That gives technical, security, and business stakeholders a shared frame for why the product exists.
Frequently Asked Questions
Concepts behind the platform
Short research notes connecting SovereignClaw to the emerging language of runtime, execution-bound AI governance.
The Execution Boundary Problem
The gap between an agent proposing an action and that action causing a side effect.
Runtime Governance for AI Agents
Governance as a dedicated runtime layer that decides authorization at execution time.
Deterministic Guardrails
Same input and policy state, same authorization outcome — reproducible by design.
Policy-Constrained Execution
Policy placed directly in the path of execution, before any adapter is reachable.