AILeash · what it costs to prove it

Everyone prices the model.
Nobody prices the proof.

At enterprise scale the model is rarely the expensive part. Ingestion, log storage, monitoring, compliance pipelines and the engineering time to hold it all together usually cost more — and none of it proves anything on its own.

Your deployment

Everything below recalculates from this.

Potential annual saving

—

—

Traditional compliance architecture —
Proof-based, on AILeash —
50p per device per month, flat

—

Year one
—
Three years
—
Per device, per year
—

Change any of these

These are the figures the saving rests on. They are illustrative, and yours will differ — so put yours in. The arithmetic follows whatever you type.

Data ingestion Getting decision data out of your systems and into somewhere it can be queried.
£
Log storage Retention at the volumes an audit trail implies, for as long as the regulation implies.
£
Monitoring platform Licences and seats on whatever watches it.
£
Compliance pipeline Turning raw logs into something a regulator will accept.
£
Engineering time Building it, and keeping it running once it exists.
£
AILeash 50p per device per month. Change it if you have been quoted something else.
£
Traditional, per device per year
—

What is measured, and what is modelled

The two halves of this page are not the same kind of number, and it matters which is which.

Measured — read from the live system just now
Blocks in the chain…
Bytes per seal32
Size of the record behind itirrelevant
Timestamp proofs confirmed…

A seal is a SHA-256 digest. Thirty-two bytes, whether the decision behind it is one line or a megabyte. That is not a claim about our architecture, it is what a hash is — and it is the whole reason the cost stops tracking the volume.

Two honest notes on the figures above. A liveness beat seals a block every five minutes, so most of that block count is heartbeat rather than customer decisions — it is not a usage number. And external timestamping is per proof: a proof is submitted first and confirmed later, so the split above is what is actually confirmed against what is still pending. Both are readable at /x/ots/status.

Modelled — assumptions, including yours
What you spend todayyour figures
What you would stop spendingan estimate

Nobody can prove what an organisation would have spent. That number does not exist anywhere to be measured, here or in any vendor's business case. What this page can do is make the assumptions visible and let you replace every one of them.

Seal this calculation

Puts your inputs and the result into the audit chain, dated and tamper-evident, and hands you a receipt anyone can check. Then what was modelled, and on whose assumptions, is a matter of record rather than of memory — including ours.

Why a proof layer costs less

It is not a discount on the same architecture. It is less architecture.

Most of that cost is moving and keeping data. Sensitive records get shipped somewhere central, held for years, indexed so they can be searched, and watched so nothing goes missing — because the plan is to reconstruct what happened by reading it all back later.

A proof-based layer answers the question at the moment the decision is made. The decision is scored, sealed into a hash chain, submitted for external timestamping and recorded by independent platforms. What survives is a proof that the decision happened, under stated rules, and has not been altered since.

So the volume stops being the problem. A seal is the same size whether the record behind it is a line or a megabyte, and it does not have to leave your systems for the proof to hold. You keep your own data where it already is.

It does not replace your logs, and it is not meant to. It replaces the machinery built to make logs trustworthy — which is the part that scales badly.

AI news
The VergeTrump’s latest land grab is an AI hub in the Philippines8mThe VergeWhat’s next for Nvidia? Whatever doesn’t rock the mind-bogglingly profitable boat19mThe DecoderGoogle Deepmind's Project Genie update brings Genie 4, chainable worlds, and a community gallery for remixing22mThe DecoderMicrosoft is replacing outside AI models in its own products with cheaper in-house alternatives1hThe VergeWhat is Nvidia without AI’s sugar daddy in charge?1hArs TechnicaWill Wright's next simulation will use AI to mine content from your memories1hTechCrunchGoogle’s AI feature in Search gets its own incognito mode1hThe DecoderOpenAI says its models delivered a year's worth of chip optimization on top of AMD's own engineering1hThe VergeDisney’s new Star Wars animatronics interact with guests using AI1hWiredInside Nvidia’s First Post-Jensen All-Hands: Pay Raises and More Equity2hTechCrunchMore banks say customers using AI agents are on their own — even if they get hacked2hThe GuardianUS senators urge White House AI chief to fix vulnerabilities after mass outages2hTechCrunchSam Altman warns against a future where AI, not humans, calls the shots2hWiredArtex Is Criminals’ New Favorite ‘Point-and-Shoot’ AI Hacking Tool3hTechCrunchOpenAI CEO Sam Altman says it’s ‘hard to imagine the AI industry without Jensen’3hThe DecoderPangram flags 99 percent of an award-winning Peruvian novel as AI-generated, but the publisher stands by its author3hThe DecoderAnthropic, OpenAI, and Google models readily marked medics, journalists, and surrendering soldiers as kamikaze drone targets4hThe VergeRemember the frontier AI pause? Chip companies don’t either4hWired‘Just Ridiculous’: A Famed Philosopher Takes on AI Consciousness4hMIT Technology ReviewThe AI Hype Index: AI is everywhere5hBBC NewsThe seven Anthropic founders who are about to become multi-billionaires5hBBC NewsHacks, AI or just overdue updates - why have there been so many outages?6hTechCrunchNvidia clarifies when it will — and won’t — keep financing neoclouds6hThe GuardianAria awards: how do Australian artists feel about their music being used to train AI? – video9hThe VergeTrump’s latest land grab is an AI hub in the Philippines8mThe VergeWhat’s next for Nvidia? Whatever doesn’t rock the mind-bogglingly profitable boat19mThe DecoderGoogle Deepmind's Project Genie update brings Genie 4, chainable worlds, and a community gallery for remixing22mThe DecoderMicrosoft is replacing outside AI models in its own products with cheaper in-house alternatives1hThe VergeWhat is Nvidia without AI’s sugar daddy in charge?1hArs TechnicaWill Wright's next simulation will use AI to mine content from your memories1hTechCrunchGoogle’s AI feature in Search gets its own incognito mode1hThe DecoderOpenAI says its models delivered a year's worth of chip optimization on top of AMD's own engineering1hThe VergeDisney’s new Star Wars animatronics interact with guests using AI1hWiredInside Nvidia’s First Post-Jensen All-Hands: Pay Raises and More Equity2hTechCrunchMore banks say customers using AI agents are on their own — even if they get hacked2hThe GuardianUS senators urge White House AI chief to fix vulnerabilities after mass outages2hTechCrunchSam Altman warns against a future where AI, not humans, calls the shots2hWiredArtex Is Criminals’ New Favorite ‘Point-and-Shoot’ AI Hacking Tool3hTechCrunchOpenAI CEO Sam Altman says it’s ‘hard to imagine the AI industry without Jensen’3hThe DecoderPangram flags 99 percent of an award-winning Peruvian novel as AI-generated, but the publisher stands by its author3hThe DecoderAnthropic, OpenAI, and Google models readily marked medics, journalists, and surrendering soldiers as kamikaze drone targets4hThe VergeRemember the frontier AI pause? Chip companies don’t either4hWired‘Just Ridiculous’: A Famed Philosopher Takes on AI Consciousness4hMIT Technology ReviewThe AI Hype Index: AI is everywhere5hBBC NewsThe seven Anthropic founders who are about to become multi-billionaires5hBBC NewsHacks, AI or just overdue updates - why have there been so many outages?6hTechCrunchNvidia clarifies when it will — and won’t — keep financing neoclouds6hThe GuardianAria awards: how do Australian artists feel about their music being used to train AI? – video9h