Open Standards
Open standards for sovereign AI. Adoptable by anyone. Owned by no one.
These aren't proprietary tools. They're open standards. We've published them, documented them, and released them under MIT license so anyone can implement them. The goal isn't lock-in — it's infrastructure that outlasts any single company.
Structured Claim Discipline
The core governance protocol. Classifies every claim as FACT, ESTIMATE, or UNKNOWN. Enforces truth-state discipline across AI responses. Prevents confident wrongness through structural design.
Identity Reflection Standard
Portable, user-sovereign identity format. The MIRROR SEED spec that allows your identity to work across any AI platform. You own the file. You control where it goes.
FACT / ESTIMATE / UNKNOWN
Every claim must declare its confidence level. FACT (verified, citeable), ESTIMATE (reasoned but uncertain), or UNKNOWN (explicit uncertainty). No more confident wrongness.
Active Mirror Governance Layer
The security gating layer. Validates requests before they reach the AI. Enforces boundaries, prevents prompt injection, maintains integrity. Red-team tested.
AMI (Active Mirror Identity)
The JSON schema for machine-readable identity. Includes human profile, communication preferences, governance rules, and continuity chain. The foundation of sovereign AI identity.
Automated Verification
A dedicated adversarial loop that scans all AI outputs for truth-state violations. If a FACT claim is made without citation, the scanner flags it. If an ESTIMATE is presented as certainty, it downgrades it.
Adversarial Correction
The system doesn't just fail; it repairs. When the Truth Scanner detects drift, the Self-Healing Loop triggers an immediate correction cycle, updating the kernel state without human intervention.
Context Synthesis
Automated pre-dawn synthesis of open loops, calendar events, and vault notes. The AI "wakes up" before you do, preparing a structured briefing document for your review.
All protocols are MIT licensed. You can implement them in your own systems, contribute improvements, or build compatible tools. The goal is interoperability, not vendor lock-in.