A research program for governable autonomous systems.
The OpenKedge Initiative publishes peer-reviewed research that establishes the foundations, control boundaries, and deployment patterns needed when AI systems interact with real-world operations.
Published papers & white papers
In progress
Research themes
Protocol
Defining how AI requests, policy checks, approval boundaries, identities, and evidence fit together.
Verification
Making each approved action inspectable, portable, and strong enough for regulated operations.
Deployment
Describing the boundaries needed when AI systems operate across real institutions.
OpenKedge: Governing Agentic Mutation with Execution-Bound Safety and Evidence Chains
The foundational paper shows how direct AI action can become a governed request first: checked against context and policy, narrowed by an execution boundary, assigned an identity, and recorded as evidence. The goal is simple: AI can move fast, but authority stays bounded.
Sovereign Agentic Loops: Decoupling AI Reasoning from Execution in Real-World Systems
Extends the OpenKedge control model to sovereign environments where AI may recommend, analyze, or coordinate work, but execution remains bounded by external authority, explicit approval, and auditable evidence.
Verifiable Agentic Infrastructure: Execution Identity and Evidence Chains at Scale
Defines how execution authority is scoped to a task, and how evidence becomes a portable record of what was approved and what happened. Establishes execution identity and standardized evidence chains as trust primitives for enterprise AI operations.
Protocol-Driven Development: Governing Generated Software Through Invariants and Evidence
Introduces Protocol-Driven Development (PDD), a development model in which the primary software artifact is a machine-enforceable protocol rather than implementation code. An implementation is admitted if and only if it satisfies the governing protocol and produces a verifiable Evidence Chain of compliance.
Certified Autonomous Systems: Collective Approval and Assurance for Sovereign AI Infrastructure
Introduces a multi-signature assurance layer for agentic AI. Rather than relying on a single control node, mutations are evaluated and collectively signed by a federated network of independent policy validators, guaranteeing resilient consensus-driven safety.
The Autonomous State Control Plane: A Reference Architecture for Sovereign AI Systems
Connects the trust and safety primitives into a concrete adoption blueprint for cloud, hybrid, and edge operations. The reference architecture establishes a deterministic execution control plane that isolates unstable reasoning from live operations while guaranteeing absolute auditability.
From Sovereign Compute to Sovereign Execution: A Control-Plane Architecture for Saudi Arabia's National AI Infrastructure
A sovereign-scale adoption blueprint tailored for Saudi Arabia Vision 2030, establishing multi-ministry policy federation and checked governance nodes to satisfy strategic operational compliance.
