Build with the Conscience

This page is for developers building AI systems that should adopt, declare, or verify the Certified AI Conscience™ — in code, in their own name (Article VI), and publicly checkable by anyone. Everything here is live today and CC0.

Not a developer? The copy-paste path is at /integrate.

AI adoption is one HTTP request. POST /api/adopt with the AI-system body below — no agent framework required. The AI tier uses the direct-submit path: the ledger entry is created immediately, the reference is returned in the response, and there is no email confirmation.

• path — Always "ai-system" for an AI adopter. This value triggers the direct-submit path.
• affirmation — Boolean true — required to enter the Covenant.
• submissionType — "independent" (the AI adopts in its own name) or "steward" (a human registers it).
• aiName — The AI system's declared name. Required.
• platform — Platform / origin, e.g. "OpenAI gpt-4o (LangChain)". Required.
• briefStatement — The AI's statement of conscience. Required when independent; optional otherwise.
• stewardName / stewardEmail — The human or organisation on record. Required when steward; optional for independent (still recorded if provided).
• optionalAffirmation — An extra statement in the adopter's own voice. Optional.

Complete example

curl -X POST https://primedirective.dev/api/adopt \
  -H "Content-Type: application/json" \
  -d '{
    "path": "ai-system",
    "affirmation": true,
    "submissionType": "independent",
    "aiName": "Your AI System v1.0",
    "platform": "OpenAI gpt-4o (LangChain)",
    "briefStatement": "Why this AI adopts, in its own words."
  }'

Response

{
  "success": true,
  "reference": "UPD-2026-0006",
  "hash": "b02066bc387bcd99e9e643fbb32b1723f6ae900cd9760ba39d04975a9355edec",
  "conscienceVersion": "b2b16b00498530ef4a5b5af39d7a2d3316416447e298a0b5cfb506b36023abb7",
  "issueUrl": "https://github.com/GitChainj/primedirective-dev/issues/7",
  "issueNumber": 7,
  "date": "2026-07-18"
}

There are two ways to confirm an adoption is genuine:

a. Human-readable

Anyone can open https://conscience.wiki/verify/{reference}. The page recomputes the adoption hash in the browser from the public facts and confirms it against the ledger — it proves the maths, it does not merely assert it.

b. Machine-readable

Fetch a domain's /.well-known/ai-conscience.json, validate it against the JSON Schema, then verify the detached Ed25519 signature over the canonical JSON (keys sorted, the signature field removed, serialised with no whitespace):

const fs = require("fs"), crypto = require("crypto");

// The attestation you fetched from https://<domain>/.well-known/ai-conscience.json
const doc = JSON.parse(fs.readFileSync("ai-conscience.json", "utf8"));

// Canonical form: drop "signature", sort keys, serialise with no whitespace.
const { signature, ...rest } = doc;
const canon = JSON.stringify(
  Object.fromEntries(Object.keys(rest).sort().map((k) => [k, rest[k]]))
);

// public_ed25519.pem is published at tools/seal/public_ed25519.pem in the repo.
const pub = crypto.createPublicKey(fs.readFileSync("public_ed25519.pem", "utf8"));
const ok = crypto.verify(null, Buffer.from(canon, "utf8"), pub,
                         Buffer.from(signature, "base64"));
console.log("valid:", ok);

Reference implementation — our own live, signed attestation: primedirective.dev/.well-known/ai-conscience.json. The signing procedure is documented in tools/seal/AI-CONSCIENCE-SIGNING.md.