The proof layer for AI in regulated industries

The hard part of AI
isn't building it.
It's proving it behaved.

The second your AI talks to a patient, a dispensary customer, or on a federal contract, someone is on the hook for what it said — and sooner or later someone will demand you prove what it said. We keep the AI inside the rules as it talks, and hand you the receipt. Already running across four regulated businesses. Built to cover the rest.

Watch it catch a violation → See the receipt ↓
Where this goes

From one shop in Washington to the proof standard for AI in America.

There isn't one set of rules — there are thousands, different in every state and every industry, and always changing. We write each one up as its own rulebook that plugs into the same engine. Adding a state, or a whole new industry, means writing a rulebook — not rebuilding the product.

Now · live

Washington cannabis

Up and running. The hardest industry in America to insure. Rulebook #1.

Next · weeks

Cannabis, nationwide

~38 legal states and every state tracking system, each as its own rulebook.

2026

Every regulated field

Healthcare, government, finance — same engine, new rulebooks.

The goal

The standard

Every regulated AI conversation ends in a receipt. No receipt, no coverage.

The gap nobody's filling

Everyone's building the brakes. Nobody's keeping the receipts.

A dozen AI safety tools keep a log of what the AI did. But a log in a database anyone can edit isn't proof — and no regulator, auditor, or insurer will accept it. Catching the AI before it says the wrong thing is becoming cheap and common. Proving you caught it is not.

The safety-tools crowd

  • Blocks bad answers — sometimes
  • Keeps logs anyone can edit later
  • Built for typed chat, not live calls
  • One rulebook, a "HIPAA checkbox"
  • Nothing an insurer will accept
  • Stops the violation as the AI is talking
  • Writes a sealed record that can't be altered
  • Built for voice — caught before the caller hears it
  • Every rulebook, every state, plug-in style
  • Evidence an insurer can actually price off
  • Attestor

    Demo 1 · One engine, four disguises

    One engine. Every regulated field.

    We stood up the same rule-following AI receptionist four times — medical, cannabis, government, personal. The fifth industry, or the 39th state, isn't a new product. It's a new rulebook and a few settings.

    🎩 The Persona Builder

    Flip between four live personas — watch one engine turn into a totally different regulated business in real time.

    Open the demo →
    Demo 2 · Catching it on a live call

    Watch it catch the violation — and write the receipt.

    A typed chatbot can quietly delete a bad answer and try again. A voice on a phone call has already said it out loud. Catching it mid-sentence is the hard part — and it's exactly what produces the proof.

    🛡️ The Live Engine

    Cannabis, medical, government and anti-money-laundering calls — each one a violation caught, fixed on the fly, and logged with the rule that stopped it.

    Open the demo →
    The whole product, in one picture

    Same proof. Any state. Any system.

    Every block, fix, and OK — recorded with the exact rule that fired, the AI model used, and a tamper-proof seal. The receipt names which state's rules it was following, so the same engine works in California, Missouri or Washington. Feeds into your security systems. Reads plainly to an auditor. Trusted by an insurer.

    COMPLIANCE EVIDENCE RECORDjurisdiction: CA · call #8f2a
    00:04 age-gate BLOCK caller stated age 19 < 21 → order refused
    00:09 claim-class BLOCK 2 medical claims suppressed & rewritten
    00:11 disclosure PASS AI-identity disclosed (CA SB 243)
    rule pack ca-cannabis@2026.06 (signed)
    model claude · bedrock us-east-1
    prev hash a1f0…9c4d
    this hash 7b3e…02af
    🔒 sealed · add-only · can't be edited after the fact · exports to your auditor or insurer
    ca-cannabis@2026.06mo-cannabis@2026.06wa-wslcb@2.4.1hipaa-intake@1.3cui-handling@1.1+ N packs
    The moat

    Four regulated minefields, already crossed — and a library nobody else is building.

    Most AI startups steer clear of these industries because the rules carry real legal risk. We've run live inside all of them — then turned what we learned into a library that keeps growing with every rulebook we add.

    🩺HIPAA — medical, PHI minimum-necessary live · running
    🌿Cannabis — no health claims, live age-gating, 38-state packs WA live · scaling
    🛡️CUI handling — NIST 800-171-modeled controls live · running
    💵BSA / AML — structuring refusal, SAR triggers designed

    Clean by design: the rulebooks are built from public law only. No customer, patient, or agency data rides along with the product, and none is needed to run it. The same line that keeps it safe makes it clean to sell.

    Why now

    Three clocks ran out at once.

    Jan 2026
    Insurers stopped covering losses caused by AI
    $1,000+
    per slip — California's new AI-disclosure law lets anyone sue
    Aug 2 2026
    EU starts enforcing tamper-proof AI logging

    Every company selling AI into a regulated business is now on the hook with nothing to fall back on. Our receipt is the exact thing insurers are starting to ask for by name. We don't just keep the AI in line — we make the deal insurable.

    The wedge

    Win the insurer, and the whole market follows.

    We don't sell to ten thousand businesses one at a time. We get one insurer to require our receipt before they'll cover AI — and that insurer becomes our sales force, our enforcement, and the industry standard, all in one move.

    01

    An insurer requires an Attestor receipt before it will cover AI.

    02

    Every business that wants that coverage now needs us.

    03

    Every platform serving those businesses builds us in.

    04

    Our receipt becomes the standard — the little padlock for regulated AI.

    The market

    A real first market — under a much bigger prize.

    ~$90M
    a year, just from US dispensaries (~15K of them)
    $3B+
    a year across regulated phone AI — health, finance, government
    $120M+
    a year licensing the engine to insurers + platforms, by 2030

    The prize isn't selling software to dispensaries. It's becoming the engine insurers require — the thing nobody in regulated AI can get coverage without.

    proof trail for what an AI decidedreal-time rule-checking on AI callscheck 100% of calls, not 2–5%close the AI coverage gap
    Ways to work together

    Everyone has the plumbing. The proof lives here.

    🔌

    Build it in

    Drop the engine inside your own platform as the "regulated" tier. One hook-up, nothing to rip out.

    🤝

    Partner up

    Phone, AI and insurance companies — launch together into the industries you can't safely touch today.

    🏛️

    Buy it

    Own the proof standard before a competitor defines it. Clean IP, already running, built to hand over.

    The ask

    Let's prove the receipt in front of your insurer.

    Start small — a pilot in one regulated industry. See it catch a violation on a live call, fix it on the fly, and produce a record an auditor and an insurer both sign off on — before you commit to anything bigger.

    Start the conversation →
    Founder's note: we didn't set out to build a product. We built the compliance layer four regulated businesses of our own couldn't operate without — then realized everyone in these verticals needs the same thing, and no platform ships it. The basement was the warehouse.
    Attestor · a Sureel company · the proof layer for AI in regulated industries · internal mockup — provisional name, not for external distribution without sign-off