How It Works
How ZK-AI Works
When machines do things — AI writing text, robots taking action, systems processing data — ZK-AI creates a receipt. That receipt proves itself using mathematics, not trust. Some receipts get an independent time-stamp. A discovery system keeps them findable even as content spreads across the internet. And a Sigil shows you exactly how strong the proof is at a glance. That is ZK-AI — proof you can verify yourself, anywhere, anytime, without asking permission from any company.
- 01
One receipt. All the way through.
The journey of every ZK-AI receipt follows five connected stages. First, a machine acts — an AI generates content, a robot performs a task, a system processes information. The moment this happens, a receipt is created containing a complete record of the action. This receipt immediately carries a Sigil, a visual indicator showing its proof strength. At the base level, mathematics alone proves the receipt is authentic — no internet, no company server, no account required. For situations where timing matters, an independent witness layer can record exactly when the receipt was created. As the content travels across the internet — through screen shots, shares, copies, and exports — a discovery system ensures the original receipt remains findable from the content itself. And throughout this journey, operators who choose rigorous independent testing can earn certification marks that signal their commitment to standards. These five stages — creation, mathematical proof, time witnessing, discovery, and certification — work together as one unified system. Each part serves a distinct purpose. Each can be verified independently. Together, they create something new: proof that travels with machine actions, outlasting the companies that created them.
- 02
What Is a ZK-SNAP Receipt?
Think of it like a digital photograph's metadata, but impossible to fake. When a machine acts — an AI writes text, a robot moves, a system processes data — ZK-AI creates a signed record of exactly what happened. The receipt contains four essential pieces: who created it, what happened, a unique fingerprint of the content, and a visual Sigil showing proof strength. The math guarantees that if anything changes — even one character — the signature breaks. You can verify it anywhere, years later, without asking the original company. The proof lives in the receipt itself, not in some database that might disappear.
- 03
Why Not Just Use Logs?
Regular logs stay trapped inside company systems — the very systems you might be questioning. Companies can edit them, lose them, or lock them behind passwords. When something goes wrong, you are stuck trusting the same people you are trying to verify. Imagine an AI makes a harmful recommendation. The company checks their logs and says 'our system worked fine.' But their logs are just text files they control. They could edit them. They could delete them. They could claim the error never happened. A ZK-AI receipt is different. It travels with the content, not inside company walls. You can verify it independently. Share it with regulators. Archive it for years. The proof exists outside any single company's control — which is exactly where proof belongs.
- 04
Proven: Math Makes It Real
Every ZK-AI receipt starts with math. The numbers either check out or they do not. This is the foundation of trust in the system. The beautiful thing? You do not need anything else. No internet connection. No company server. No account that might get deleted. You can verify a Proven receipt on a computer in a basement, on a plane, or twenty years from now. The signature, structure, and content all verify from the file itself. Mathematics do not expire. They do not get acquired by other companies. They do not have outages. When everything matches, you see the Proven Sigil — your guarantee that this receipt is real, independent of anyone's word.
- 05
Recorded: Time You Can Trust
Proven means the math works. Recorded means the math works AND an independent witness wrote down exactly when it happened. Why does timing matter? Imagine a contract dispute where both sides claim they wrote the document first. Or an AI company claiming their model was trained before a certain dataset was released. Or a journalist proving they had a story before it was officially announced. The Recorded level adds an independent time-stamp from a witness layer that does not belong to any single company. It writes down: this receipt existed at this exact moment. You still verify the math yourself — the witness just confirms when. Both Proven and Recorded are valid. Use Proven when you need portable verification anywhere. Use Recorded when the exact timing matters for your situation.
- 06
Certified: Tested, Not Claimed
Anyone can use ZK-AI receipts. The protocol is open. But some operators choose to go above and beyond. These operators submit to rigorous independent testing by third-party auditors. The testing checks four dimensions: are their systems cryptographically sound, do they operate transparently, do they provide proper finality, and do they maintain appropriate context about what they record? Those who pass earn the 3DVC certification mark — like a safety inspection sticker on an elevator or an organic label on food. It proves someone impartial checked their work against published standards. The mark sits next to the Sigil, never replaces it. The Sigil shows the proof is valid. The 3DVC mark shows the operator was tested. Together, they give you two different kinds of confidence.
The Trust Flow
The Trust Pipeline
Every receipt follows the same journey through five stages. It starts when a machine does something — writes text, makes a decision, processes data. Immediately, a receipt is born. Then mathematics proves it is real, no internet required. Optionally, an independent witness records exactly when it happened. A discovery system keeps receipts findable no matter where the content travels. And some operators earn certification through rigorous third-party testing. The trust builds step by step — and you can verify every stage yourself, without trusting any company.
- Something happens
- Every machine action creates a signed receipt at the exact moment it occurs.
- Math proves it (S4)
- The receipt checks out using mathematics alone — no internet needed, no company required. This is the 'Proven' state.
- Time is recorded (S5)
- An independent witness layer records when the receipt was created, so anyone can verify both what happened and when.
- Certification (3DVC)
- Some operators go through independent testing to earn a certification mark — like a quality seal that proves they meet published standards.
- Find it anywhere
- A discovery layer keeps receipts findable even after files are copied, moved, or shared across different platforms.
Step 1
What Is a ZK-SNAP Receipt?
A receipt is a signed record of what a machine did. It could be an AI writing text, a robot taking action, or a system processing data. Unlike regular logs that live inside company databases, this receipt proves itself through mathematics.
Each receipt contains four essential elements working together: a signature proving who created it, a record of what happened, a unique fingerprint of the content that changes if even one character is altered, and a Sigil showing the proof strength. You never need to ask the original company if it is real — the math tells you instantly.
- Signature
- Who created this receipt. If anything changes, the signature breaks — that is how you know.
- What happened
- The action, content, and context — all bound together in one record.
- Fingerprint
- A unique seal proving the content has not been altered, even by one character.
- Sigil
- A visual indicator showing proof strength — from basic (Proven) to time-stamped (Recorded).
Step 2
Why Not Just Use Logs?
Regular logs stay trapped inside company systems — the very systems you might be questioning. Companies can edit them, lose them, or lock them behind passwords. When something goes wrong, you are stuck trusting the same people you are trying to verify.
Consider what happens when an AI system makes a harmful recommendation, then the company claims their logs show "everything worked fine." But those logs are just files on their servers. They could have been modified. Deleted. Or never recorded accurately in the first place. You have no way to check.
A ZK-AI receipt is different. It travels with the content itself, not inside company walls. You can verify it independently using only mathematics. Share it with regulators. Archive it for decades. The proof exists outside any single company's control — which is exactly where proof belongs in a world of automated systems.
Step 3
Proven: It Checks Out Anywhere
Every ZK-AI receipt starts with mathematics. The numbers either check out or they do not — there is no middle ground. This foundation is what makes the entire system trustworthy.
Here is the powerful part: you need nothing else. No internet connection. No company server that might go down. No account that could be deleted. You can verify a Proven receipt on a computer in a basement, on an airplane with no wifi, or twenty years from now when the original company may not even exist anymore.
Mathematics do not expire. They do not get acquired by other companies. They do not have outages or maintenance windows. The signature, structure, and content all verify directly from the receipt file itself. When everything matches, you see the Proven Sigil — your absolute guarantee that this receipt is authentic, independent of anyone's word or any company's continued existence.
Step 4
S5 Recorded: Witnessed in Time
Two levels of proof. Both valid. Just suited for different situations. Proven means the math checks out anywhere, anytime — no internet connection required. Recorded adds an independent time-stamp from a witness layer that belongs to no single company, so you know exactly when the receipt was created.
Timing matters more than you might think. Imagine a contract dispute where both sides claim they wrote a document first. Or an AI company claiming their model was trained before a harmful dataset was released. Or a journalist proving they had a story before it was officially announced. In all these cases, the exact moment matters as much as the content itself.
The Recorded level adds this independent time verification while keeping everything else the same. You still verify the math yourself — the witness layer simply confirms "this receipt existed at this exact moment in time." Use Proven when you need portable verification that works anywhere. Use Recorded when establishing the precise timing is important for your situation.
Step 5
The Evidence Graph
In the real world, people share content — images, documents, recordings, screen shots — not receipt files. A photo gets posted to social media. A document gets emailed around. A video gets re-uploaded to different platforms. The original receipt file often gets lost in the shuffle.
So ZK-AI built a discovery system that works backwards from the content itself. You upload a screen shot, a copy, or an export. The system analyzes the content and finds any receipts associated with it, then shows you the proof of what originally happened. The content becomes the key to finding its own history.
Finding receipts and verifying them are separate jobs by design. The discovery system helps you locate receipts across the network. The verification system proves they are authentic using mathematics. One helps you search, the other proves what you found is real. This separation ensures that no single system controls both discovery and truth.
receipt_idbatch_rootanchor_id- status
- time
- source artifact
- dataset
- audit package
- claim details
- sealed evidence
Step 6
Certified: Extra Testing, Extra Trust
Anyone can create and use ZK-AI receipts. The protocol is intentionally open and accessible. But some operators choose to go further — submitting to rigorous independent testing by third-party auditors who examine their systems against published standards.
These auditors check four critical dimensions: cryptographic soundness, operational transparency, proper finality of records, and appropriate context maintenance. It is thorough, impartial, and designed to ensure operators meet high standards.
Those who pass earn the 3DVC certification mark — similar to a safety inspection sticker on an elevator or an organic certification on food. The mark sits next to the Sigil, never replaces it. The Sigil shows the proof itself is valid. The 3DVC mark shows the operator was independently tested. Together, they give you two complementary kinds of confidence: one in the mathematics, one in the organization behind it.
- Sigils
- Automatic proof indicators that show up on every receipt. Like a battery meter, but for trust.
- 3DVC Mark
- An extra certification earned through independent testing — like a safety inspection sticker.
- The difference
- Everyone can use receipts and show Sigils. Only tested operators earn the 3DVC mark.
Applications
What ZK-SNAP Is For
- Creators and rights
- Artists, musicians, and writers can prove exactly when they created something and what version is authentic. No more he-said-she-said about who made what first. The receipt travels with the file, so proof stays with the work even as it gets shared across platforms.
- AI and data transparency
- When an AI generates content, a receipt can record what training data shaped its response. Companies can prove their models were trained ethically. Users can verify an AI's claims about its sources. The black box becomes checkable.
- Automated systems
- Factories, drones, and autonomous vehicles can create receipts for every action they take. If a robot makes a mistake, the receipt shows exactly what happened, under what safety protocols, and when. Not for blame — for learning and accountability.
- Public records and journalism
- Governments can prove documents are authentic without revealing sensitive details. Journalists can verify leaked materials came from specific sources. Whistleblowers can prove their claims without exposing their identity. Proof and privacy, together.