Skip to main content

BASIS

Baseline Authority for Safe & Interoperable Systems

The open standard for AI agent governance


The Problem

AI agents are making autonomous decisions. Right now, there's no standard way to:

  • Verify an agent will behave within bounds
  • Trust that governance checks happen before action
  • Audit what decisions were made and why
  • Interoperate between different agent systems

Every company is building their own governance. None of it talks to each other. None of it is verifiable.


The Solution

BASIS is an open standard that defines what must happen before an AI agent acts.

┌─────────────────────────────────────────────────────────────┐
│ AGENT WANTS TO ACT │
└─────────────────────────────────────────┬───────────────────┘


┌──────────────────────────────────────────────────────────────────────────┐
│ BASIS GOVERNANCE │
│ │
│ ┌───────┐ ┌─────────┐ ┌─────────┐ ┌─────────┐ ┌─────────┐ │
│ │ CAR │──▶│ INTENT │──▶│ ENFORCE │──▶│ PROOF │──▶│ CHAIN │ │
│ │ │ │ │ │ │ │ │ │(optional)│ │
│ │Identity│ │ Parse │ │ Check │ │ Log │ │ Anchor │ │
│ │Resolve│ │ Plan │ │ Trust │ │ Audit │ │ Verify │ │
│ │Creds │ │ Risk │ │ Gate │ │ Trail │ │ Immutable│ │
│ └───────┘ └─────────┘ └─────────┘ └─────────┘ └─────────┘ │
│ │
└──────────────────────────────────────────┬───────────────────────────────┘


┌─────────────────────────┐
│ ALLOWED │ DENIED │
│ Execute │ Block │
└─────────────────────────┘

The Five Stages

StagePurposeKey Question
CARResolve agent identity and credentials"Who is this agent and what can it do?"
INTENTUnderstand what the agent wants to do"What is being attempted?"
ENFORCECheck if it's allowed based on trust & policy"Should this be permitted?"
PROOFCreate immutable audit trail"What happened and why?"
CHAIN (optional)Anchor proofs to blockchain"Can this be independently verified?"

Core Principles

1. Governance Before Execution

No autonomous action proceeds without passing through governance checks. Period.

2. Trust is Quantified

Not binary allow/deny, but graduated trust scores (0-1000) that unlock capabilities progressively.

3. Everything is Auditable

Every governance decision is logged with enough detail to reconstruct exactly what happened.

4. Open Standard, Many Implementations

BASIS is the spec. Anyone can build a compliant implementation. No vendor lock-in.


Trust Tiers

TierScoreCapabilities
T0 Sandbox0-199Sandbox only
T1 Observed200-349Limited operations
T2 Provisional350-499Basic operations
T3 Monitored500-649Standard operations
T4 Standard650-799Extended operations
T5 Trusted800-875Elevated operations
T6 Certified876-950Privileged operations
T7 Autonomous951-1000Full autonomy

Why Open?

For Developers:

  • Build once, deploy anywhere
  • No proprietary lock-in
  • Community-driven improvements

For Enterprises:

  • Vendor-neutral standard
  • Auditable compliance
  • Interoperable agents

For the Ecosystem:

  • Shared infrastructure costs
  • Network effects for trust
  • Rising tide lifts all boats


Implementations

Cognigate (Reference)

The reference implementation maintained by Vorion.

Build Your Own

Anyone can implement BASIS. See the implementation guide.


Get Involved

BASIS is community-governed. Here's how to participate:

  1. Use it — Build agents on BASIS
  2. Contribute — Code, docs, ideas
  3. Govern — Shape the standard's future

Join the Community →


Status

ComponentStatus
Specificationv1.2 (Draft)
Reference ImplAlpha
Test SuiteIn Development
CertificationComing Soon

BASIS is an open standard released under CC BY 4.0. Reference implementations are Apache 2.0.