Monop Content presents AILeash — a compliance API for platforms running AI. Governance, child safety and fraud alerts on one tamper-evident engine

Every AI decision.
Sealed. Provable. Yours.

The compliance layer you build on top of. You put sebbi.pro underneath your app; it scores every AI decision in under 30ms — ALLOW, CHALLENGE or BLOCK — and seals each one into a SHA-256 chain nobody can quietly edit. Not a hacker. Not an employee. Not even us — the chain is timestamped through OpenTimestamps into Bitcoin, so the clock sits outside our control, and other platforms witness our chain so we cannot rebuild it either. Built to support EU AI Act Articles 9, 12, 13 and 14, the Online Safety Act, the ICO Children's Code and the DSA. When a regulator asks what your AI decided and why, you answer in one API call. Free for 90 days. Then 50p per device. No tiers, no sales calls.

Three steps. That is it.

What normally costs £50,000 a year in compliance tooling, free for 90 days and then 50p per device.

Add one line. Never think about it again.

The slowest part of any compliance tool is wiring it in. So we removed that part. Put one line above a function and every call it makes from then on is fingerprinted and sealed — automatically, forever, with nothing else to remember.

Five products. One engine.

The same 9-signal engine and audit chain underneath all of them. Pick one or take the lot. Every one is free for your first 90 days. Four run on our infrastructure; Sebdog runs on yours.

The data never leaves. The proof still does.

Every compliance vendor asks you to send them your data. That is a straight trade: you get an audit trail, they get your records. For a hospital, a council or a defence supplier that trade is simply not available, so those organisations end up with nothing but a spreadsheet.

Sebdog is the same engine running inside your building. Your decisions, your events and your hash chain sit on your disk and never move. We cannot read them. We cannot subpoena them out of our own systems, because they were never in our systems.

Which leaves the obvious problem, and it is the one everybody skips: a hash chain on your own server proves nothing against you. You own the file. You own the keys. You can rebuild it forward, re-sign it and renumber it, and nobody outside can tell. An audit trail the owner can silently rewrite is a diary, not evidence.

Nine signals as standard. Then add your own.

Every event is scored against the nine core signals below. A Signal Pack loads extra signals on top, tuned to one kind of risk — and the pack itself is sealed into the chain, so the rules that were running at the moment of a decision are part of the record, not a memory.

You set the price. You keep the rest.

Your first 90 days are free. After that we take 50p per device per month. Everything above it is yours — every month, for as long as they stay.

Tell a friend. Earn forever.

Your signup comes with a referral code. Every device that signs up with it pays you 10p a month, for as long as it stays. No cap, no expiry.

Seal your profile before someone clones it.

AI can now copy a face, a voice and a bio in minutes. The Sovereign Profile Notary locks your public identity — name, bio, links — into the same tamper-evident chain that seals AI decisions, with an official timestamp. You get a short verification code for your LinkedIn bio; anyone can check it in seconds. If an impersonator changes even one letter of your profile, the check fails in public. Free, takes 60 seconds, and your details never leave your browser unless you choose to publish them.

And for businesses: the Payment Notary. Invoice fraud costs UK businesses hundreds of millions a year — criminals intercept real invoices and switch the bank details. Seal your true details once, print a short code on every invoice, and every customer can verify in 10 seconds before paying. Switched details fail the check — the fraud dies before the money moves.

What the law actually requires.

Most compliance advice tells you what the text says. This is what it means for your platform in practice — including the dates that moved.

Article by article. Feature by feature.

The law names its requirements. Here is the exact engine feature built to answer each one — taken directly from the regulation map sealed inside the engine itself. No hand-waving: requirement on the left, mechanism on the right.

Nobody can prove a person thought about it.

We will say the thing the rest of this market avoids. You cannot prove genuine human oversight happened. Thinking is an internal state and no amount of logging reaches it. Any vendor telling you they have solved that is selling you something that does not exist.

But rubber stamping is not an internal state. It is a pattern — and patterns leave marks, if you record the right things in the right order at the time. That part is solvable, and this is how.

Pinned from both sides. Before and after the fact.

A hash chain proves nothing was altered after the fact. On its own it does not prove when the chain was built — because in principle whoever runs the system could rebuild the whole thing and date it however they liked. Every vendor promising you an immutable audit trail has this hole in it. Most don't mention it.

So we moved the clock outside our own building. At regular intervals the current tip of the chain is submitted to OpenTimestamps, which aggregates it with thousands of unrelated timestamps into a single Merkle root and commits that root to the Bitcoin blockchain. Several independent calendar servers do this in parallel, so no single one of them can fail, disappear or lie without the others contradicting it. From that moment the timestamp is not ours to move, not ours to re-issue, and not ours to quietly correct. Rewriting a sealed record would mean rewriting Bitcoin.

Said precisely, because a reviewer found us saying it loosely. Anchoring is a property of an individual proof, not of the chain. Submitting a tip to OpenTimestamps is not the same as that proof being confirmed in a Bitcoin block — confirmation takes hours, and until it lands the proof is pending. So a tip is either covered by a confirmed proof or it is not, and we will not describe the chain as anchored while a proof covering it is still waiting.

Check the real state at /x/ots/status — it lists every proof, which are confirmed and which are pending, and returns the raw .ots file so you can verify it against Bitcoin without us. We publish that route because it is the one that can show us waiting.

One chain can be rebuilt. Ten cannot.

A confirmed anchor stops us backdating anything older than it. Good — and still not the end of it. So platforms witness each other.

This is running, not planned. Four chains are live and exchanging tips right now, one of them unattended and hourly since the first of August with several hundred observations behind it. The list is public and machine-readable at /x/roster/list — no key, no account, no permission. Peers run a single file on a cron line; there is nothing to install and no code of ours inside their stack.

Each platform periodically hands its current chain tip to its peers, and each peer seals that tip into its own chain. From that moment your history sits inside chains you do not control — chains that are themselves anchored externally. To rewrite your own past you would now need every peer who witnessed you to rewrite theirs in step and re-anchor all of it. That stops being a technical exercise and becomes a conspiracy, and it gets harder with every platform that joins.

Three ways to submit. Each says a different thing.

A peer chooses the lane. Every lane is free, and the record says which one was used, so nobody's submission is described as stronger than it was.

Sealing the things that get argued about.

The engine seals decisions. These seal the processes around them — the parts that turn into a dispute eighteen months later, when everyone's memory has conveniently improved. Each one writes into the same chain, so the same verification and the same anchor cover all of it.

The thing you actually hand to the auditor.

Everything above produces a chain. A chain is not a deliverable. When a regulator, an insurer, a court or an enterprise customer asks you to prove it, you need one file you can send them — and it has to survive being read by someone whose job is to find the hole in it.

So the pack does not summarise the chain. It re-verifies it. Every block is rewalked and rehashed from genesis at the moment the pack is built. A report tells you what a system claims about itself; this recomputes the arithmetic and fails loudly if it doesn't come out. Those are not the same product, and only one of them is worth anything in a dispute.

The instrument that would catch us failing.

Three things on this page have a soft edge, and we would rather name them than wait for you to find them. Oversight sealing only works if the platform using it doesn't show reviewers the verdict early. Witnessing only means something with breadth. A declaration is only as strong as the rules in it.

None of those can be fixed by arithmetic. All three can be measured — and a weakness somebody is measuring is a very different thing from one nobody is.

Don't take any of it on trust.

Everything above this line is a claim. Below it is the live engine — no account, no email, no key. Send it a case and it scores it, seals it into the production chain, tells you exactly what would have changed the answer, and hands you the block number so you can check the lot yourself at endpoints that have never heard of you.

Every case run here becomes a genuine block in the production chain, covered by the same external timestamp as every customer decision. Public endpoints are rate limited, and the arithmetic is published in the whitepaper if you would rather check it than run it.

Authority, derived. Not assumed.

Everything above proves what your AI did. This proves it was entitled to. An agent takes an action; it got its authority from another agent, which got it from a system, which got it from a person. Nothing in the standard governance stack can show the authority used at the end derives from the authority granted at the start. Permissions answer one hop. Audit logs describe the aftermath. Neither derives anything.

Five minutes. Start to sealed.

No build step, no container, no dependencies to install, nothing added to your requirements file. If your organisation reviews new libraries before they go near production, this is one readable file rather than a supply chain.

AILeash API key. 90 days free.

Pick a product, add your name and email. Key, referral code and a step-by-step installation guide arrive instantly — everything free for 90 days, no card. When the trial ends you'll be directed to a secure Stripe payment reflecting only the real devices that used your key.

Your API key — save it now

Your referral code

Share it with anyone. Every device they sign up pays you 10p a month, for as long as it stays.

POST https://sebbi.pro/api/govern
Authorization: Bearer YOUR_KEY
Free for 90 days · 50p per device after the trial · billed via Stripe on real devices only

Your shield — live-verified, put it on your site

This badge is drawn live by our server. It shows gold and your company name only while your account is active — copied or faked, it renders grey UNVERIFIED.