← COHIBA

Security Evidence

This page publishes reproducible build evidence. It does not claim an independent audit.

Verification model

Fixed supply, destination balance, immutable metadata, mint authority revocation and freeze authority revocation.

Independent audit status

NOT YET COMPLETED — REQUIRED BEFORE MAINNET.

Human Signal — PRE-AUDIT

Experimental authorization and provenance for AI Agents. Independent security audit, a live issuer, independently operated external verification and a documented public adversarial round are all outstanding. Passing tests does not establish safety.

Four release gates: BLOCKED. Evidence unavailable does not clear a gate.

Current release readiness · Audit scope · Testing scope

Current privacy claim: data minimization and separation of private evidence. Stable identifiers remain linkable; anonymity is not implemented.

Agent Control research pilot

A local reference gateway verifies PoHA, enforces shared call/byte budgets and rechecks policy before two harmless tools: synthetic-data reading and local draft storage. Queued actions can be cancelled by operator suspension. This pilot is not activated on this website and does not establish AI or superintelligence containment.

Source, reproduction and control assumptions

Build evidence

Loading…
ArtifactSHA-256

Specifications

Verification Specification · Threat Model · Security Control Matrix · Security Policy