What did GeoClear just launch?
The short version, the slightly longer version, and answers to the questions we keep getting. No jargon, promise.
Imagine you mail a letter without a stamp. Nobody can verify who sent it, when, or that it wasn't tampered with along the way. The letter is "naked."
AI agents today are sending "letters" all day long. Answers, decisions, transactions, approvals. None of them carry stamps.
We built the stamp. Every GeoClear answer comes with an invisible sealed operational evidence record that verifies who asked, when, what was answered, and that nothing's been changed. The verification is deterministic and you can run it yourself.
Think of it as a flight recorder for AI agents. When something goes wrong (and it will), you have the replay.
Isn't this just an authentication standard?
No. A standard gets an agent through the front door with a credential. GeoClear verifies whether each action that credential attempts is allowed, and hands you a record you can check yourself. Holding the credential is not the same as the action being approved.
If the model already refuses harmful requests, who needs an action verifier?
Model safeguards govern what the model writes. They do not give your receiving system, your auditor, or your regulator an independent record that a specific action followed the approved path. That record is what GeoClear holds.
Can you explain it like I'm five?
Imagine you mail a letter. Without a stamp from the post office, no one believes when you sent it or who sent it. The letter is "naked." Anyone could fake it.
Today's AI agents are like that. They send "letters" (answers) all day, to other agents, to apps, to companies. None of those letters have stamps. Nobody can verify later who sent it, when, or what the question even was.
We built the stamp.
Every time an AI agent takes a high-impact action (like "approve this transaction" or "escalate this claim"), GeoClear can put a tiny invisible seal on the record of that decision. The seal says: who asked the question (so the answer can't be swapped onto a different question later), when (down to the second), who issued it (us, and only we can make the seal, because the key is locked in a special chip), and what the answer was (so it can't be edited after).
Anyone can check the seal, like reading a UPC barcode. They don't have to trust us; the math just works out, or it doesn't.
Why does this matter? Because in a few years, all of it (your bank, your insurance, your delivery, your lawyer's paperwork) will be done by AI agents talking to other AI agents. When something goes wrong (and it will), someone will ask: "Show me the answer the AI used." Today, there's nothing to show. We just made the thing to show.
What is GeoClear, in one sentence?
The independent operational evidence layer for AI actions: at the moment intent becomes an action, GeoClear verifies it against the approved policy and issues a tamper-evident operational evidence record the customer holds and anyone can verify later, offline.
What does "verification over information" mean?
Most APIs sell information. They give you an answer, you trust the vendor. We sell verification. We give you an answer plus the evidence that it's real, verified, and replayable. If the answer is ever questioned, you don't have to call us. You just hand over the record.
Why does this matter now?
Because AI agents are about to be doing real work. Booking flights, approving loans, signing contracts, dispatching drones, running compliance checks. When something goes wrong, somebody (a regulator, an auditor, a customer) will ask: "Show me the answer the AI used on March 14." Today, there's nothing to show. We made the thing to show.
What's actually in the operational evidence record?
Four things, sealed together: the question that was asked, the answer that was given, the time down to the second, and a signature from a key held inside a hardware-isolated signing boundary. The signature can only be made by us, but anyone can check it without contacting us, and any modification to the sealed answer is tamper-evident on verification.
How do I check an operational evidence record myself?
Use the offline verifier to confirm what GeoClear returned at decision time, or check the X-GeoClear-Receipt header on any response from geoclear.io/api/health. The operational evidence records are designed for independent verification with customer-held verification material, no GeoClear server is required for the verification result. Developers building a verification step into their own pipeline can find the public verifier package in the developer docs.
Is this blockchain?
No. Same family of math (cryptographic signatures), no blockchain, no token, no crypto coin. Just a sealed answer on a regular API call.
Is this AI?
No. We don't build AI. We build the thing that makes other people's AI safe to use in places where the wrong answer costs real money. Think seatbelts: we don't make cars; we make the safety part.
Who buys this?
Teams deploying AI agents into regulated and mission workflows who need evidence before accepting a high-impact action: business leaders who need agents to act without unchecked authority, CISOs who need policy-bound records enforced at the edge, and risk, legal and compliance teams who need evidence that the approved path was followed, not just a claim that the AI was right. See evidence profiles for which capabilities are shipped today versus product direction.
What does it cost?
GeoClear is licensed to an enterprise and deployed inside your own cloud account. There is no self-serve tier and no subscription. It is bought in one of two ways: a paid evaluation, a fixed fee for a set number of transactions in one real workflow, or a negotiated enterprise or federal contract. Pricing is set for each engagement rather than published as a list, and the licensing page sets out what determines it and how to get a number.
How is this different from SmartyStreets / Google / Loqate?
They give you an answer. We give you an answer plus a tamper-evident operational evidence record that verifies what was returned at decision time. Same way "a photo" is different from "a notarized photo with a date stamp." None of them carry customer-held operational evidence records on every response, and adding that pattern is not a copy edit; it requires building the trust boundary into the substrate, not bolting it on at the response layer.
Is the operational evidence record system live today?
Yes. Every JSON response from geoclear.io has carried a tamper-evident operational evidence record since 2026-04-23. Reviewers verify the record independently with customer-held verification material, see the offline verifier or the developer docs for the public verifier package.
Where do I see it work?
Live demo on the homepage at geoclear.io. The offline verifier at geoclear.io/security/offline-verifier shows the full operational evidence record with a click-to-verify flow.
Where can I read the deeper thesis?
Two pieces, depending on what kind of reader you are:
📜 The thesis: The Machine's One-Word Reply: Why JSON Isn't Enough. Why the agentic economy needs tamper-evident operational evidence records. Read the Manifesto →
📡 The product launch: the 2026-04-30 launch post on Sovereign Trust about the GeoClear operational evidence record substrate. What we shipped, why now, and how to verify an operational evidence record yourself. Read on Sovereign Trust →