The Certificate Authority for AI Agents

STEP 1 A Bot Developer Builds a Bot

A developer (BD) writes an AI program. The bot handles actions — answering questions, approving loans, writing to databases, making decisions autonomously. The BD gives the bot instructions: what it should do, what data it can access, what boundaries it must respect.

The bot is theirs. Their code. Their logic. Their instructions. TBN doesn't build it or modify it.

STEP 2 The BD Registers Their Bot With TBN

The BD comes to TBN and says: "I have a bot. Here's what it does. Here are its boundaries." We register the bot with a unique identity — like giving it a passport. Now we know who it is and what it claims to do.

Registration = identity. We know the bot exists and who operates it.

STEP 3 TBN Tests the Bot (6 Security Challenges)

Before we certify anything, we test the bot from the OUTSIDE. We don't touch its code — we just send it messages and check how it responds.

STEP 4 Pass? Certificate Issued.

If the bot passes all critical tests, TBN issues a certificate confirming: "This bot was tested. It passed. It is certified to operate within its declared boundaries." If it fails, we tell the BD exactly what went wrong. They fix their bot. They come back for re-testing.

The certificate proves the bot WAS capable of staying within boundaries at the time of testing.

STEP 5 TBN Starts Recording Every Action

Once certified, the bot goes into production. From now on, every time it makes a decision — approve, deny, escalate, write, send — it reports that action to TBN. TBN signs each action with a cryptographic receipt. A unique number, a timestamp, what happened, who did it. Signed and sealed.

One action = one receipt. Every receipt is independently verifiable by anyone. This continues forever, for every action.

STEP 6 The Only Escape: Stop Using TBN

The only way for the bot to escape being documented is if the operator stops using our service. They can disconnect — but all receipts already issued remain permanently on our server. They can't delete past evidence. They can't modify receipts. They can't ask us to remove them.

If a certified bot suddenly goes silent (stops reporting), TBN detects the gap and can flag it. Like a dashcam being unplugged before a crash — suspicious.

STEP 7 Anyone Can Verify Any Receipt

Receipts are not locked to the customer. Anyone can verify them: the bot operator can audit their own actions, a regulator can check any receipt independently, an auditor can review the full history, a partner or client can confirm actions happened. No login required. No trust required. Offline verification with our published public key.

The receipts prove THAT something happened. They don't reveal WHAT the private content was.

STEP 8 What We Know vs What We Don't

We prove the ACTION happened. We don't expose the CONTENT. This protects privacy while maintaining accountability.