Protocols over Platforms
We believe that for AI to be safe, its governance must be public, peer-reviewed, and mathematically verifiable.
SCD Protocol v3.1
Sovereign Compact Directive
The SCD is a formal protocol for defining the "Constitution" of a local AI system. It enforces truth-state classification (FACT/ESTIMATE/UNKNOWN) at the kernel level, ensuring that model outputs are strictly bounded by verifiable data.
Desai, P. (2025). The Sovereign Compact Directive: A Protocol for Truth-State Enforcement in Local AI. Zenodo. https://doi.org/10.5281/zenodo.17787619
Publications & Preprints
Governance and Boundary Conditions for Reflective AI Systems
aiXiv Preprint · Dec 2025
Proposes a novel architecture where "Reflection" is not just a prompting strategy, but a separate, adversarial node in the system topology.
Layered Governance for Large Language Model Systems
aiXiv Preprint · Dec 2025
A framework for "Governance as Code," moving safety checks from the model weights (RLHF) to the orchestration layer (Runtime).
Our Validation Philosophy
1. Public Proof
All safety claims must be backed by reproducible code and public datasets. No "black box" safety.
2. Adversarial Design
Systems are secure only when they survive dedicated red-teaming. Our "M1 Adversary" node is proof of this commitment.