VERDICT WEIGHT

A confidence scoring framework for autonomous AI systems. USPTO-protected, IEEE-validated, defense-ready.

What this is

VERDICT WEIGHT™ is a confidence scoring framework that lets autonomous AI systems quantify, audit, and defend the trustworthiness of their own decisions. It composes eight independent evidence streams into a single, calibrated confidence score — with cryptographic provenance and a registry-level kill switch for adversarial conditions.

It is built for environments where AI decisions carry real consequence: defense, critical infrastructure, regulated industry, and any deployment where “the model said so” is not an acceptable audit trail.

Get Started

Install in one command and score your first decision.

Architecture

The 8-stream composition and why each one is necessary.

Validation

Head-to-head results vs. Dempster-Shafer, Naive Bayes, and averaging baselines.

Defense Positioning

Why VERDICT WEIGHT is built for adversarial autonomy.

Status

USPTO

Provisional Patent #64/032,606 · Trademark #99747827

IEEE-grade Validation

673/673 tests · 27 suites · 7 reviewer attack categories addressed

Distribution

PyPI · GitHub · Zenodo (DOI: 10.5281/zenodo.19447547)

Who runs this

VERDICT WEIGHT is developed by Odingard Security, a division of Six Sense Enterprise Services LLC, founded by Andre Byrd. For pilot inquiries, contact [e-mail].