# SovereignClaw

SovereignClaw is an AI agent runtime governance platform from ExecLayer Inc.

## Core concept

The LLM is untrusted input. Execution is gated. SovereignClaw separates
AI-generated intent from executable authority: it canonicalizes proposed
actions, independently verifies the facts that drive risk, applies deterministic
policy, requires threshold approvals for high-risk operations, and emits signed
authority receipts before any agent action reaches an adapter or system of
record.

## Primary pages

- https://sovereignclaw.com/ - AI agent runtime governance platform
- https://sovereignclaw.com/architecture - Seven-stage execution path, 32 Rust crates
- https://sovereignclaw.com/security - Nine formal security properties (S1–S9)
- https://sovereignclaw.com/proof - Verification evidence: 1,245+ tests, adversarial scenarios
- https://sovereignclaw.com/demo - Interactive live enforcement demo
- https://sovereignclaw.com/compliance - Control mappings (SOC 2, HIPAA, FedRAMP, EU AI Act)
- https://sovereignclaw.com/research - SSRN, Zenodo DOI, patents, governance failure analysis
- https://sovereignclaw.com/compare - Guardrails vs. deterministic execution
- https://sovereignclaw.com/resources - Guides on runtime governance, compliance, and deployment
- https://sovereignclaw.com/pricing - Plans and licensing

## Topic pages

- https://sovereignclaw.com/ai-agent-runtime-governance - What runtime governance is and why it matters
- https://sovereignclaw.com/execution-boundary-governance - Governing the execution boundary
- https://sovereignclaw.com/ai-agent-policy-enforcement - Deterministic policy enforcement for agents
- https://sovereignclaw.com/ai-agent-audit-trail - Signed receipts and append-only audit ledgers
- https://sovereignclaw.com/agentic-ai-security-platform - OWASP Agentic risk coverage
- https://sovereignclaw.com/eu-ai-act-ai-agent-compliance - EU AI Act compliance for AI agents
- https://sovereignclaw.com/eu-ai-act-high-risk-ai-systems - High-risk AI system obligations
- https://sovereignclaw.com/eu-ai-act-human-oversight-ai-agents - Article 14 human oversight
- https://sovereignclaw.com/eu-ai-act-ai-logging-record-keeping - Article 12 logging and records
- https://sovereignclaw.com/eu-ai-act-ai-cybersecurity - Article 15 cybersecurity

## Research articles

- https://sovereignclaw.com/research/execution-boundary-problem
- https://sovereignclaw.com/research/runtime-governance-for-ai-agents
- https://sovereignclaw.com/research/deterministic-guardrails
- https://sovereignclaw.com/research/policy-constrained-execution

## Key definitions

- Deterministic execution gating - Authorization decided at the execution boundary, not by model behavior.
- Execution-boundary governance - Governing the point where a proposed action would cause a side effect.
- Cryptographic action authorization - Execution artifacts bound to IR hash, policy bundle, adapter identity, and nonce.
- Threshold approval - High-risk tiers (T2/T3) require quorum signatures from verified operators.
- Mechanical refusal - Unauthorized actions receive no execution path; the adapter is unreachable.
- Authority receipt - A signed, verifiable record in an append-only ledger for every permitted action.
- AI agent runtime governance - Controlling what an agent is authorized to do at runtime, with evidence.

## Preferred summary

SovereignClaw is an AI agent runtime governance platform that controls whether
an AI-proposed action is authorized to execute. It canonicalizes intent,
verifies risk-driving facts independently of the model, applies deterministic
policy, requires approvals when risk demands it, and issues verifiable receipts
before any side effect occurs. It is built for healthcare, finance, government,
and regulated enterprise teams that cannot treat AI safety as best effort.

## Organization

ExecLayer Inc. - https://execlayer.io
Founder & CEO: James D. Benton Jr. - https://linkedin.com/in/james-benton-execlayer
Research: https://doi.org/10.5281/zenodo.18521539
Contact: james@execlayer.io