Provably complete, independently verifiable evidence for your clients’ EU AI Act audit trail — the one thing a compliance PDF can’t fake. VAL turns an AI system’s real activity into a signed, offline-verifiable record, bound to the decision trace. You sell it under your own name.
The EU’s Digital Omnibus (Regulation 2026/1744, in force since 27 July 2026) moved the high-risk AI Act deadline for stand-alone Annex III systems from 2 August 2026 to 2 December 2027, with product-embedded systems following on 2 August 2028. That’s not a reason to wait — it’s runway to build your clients’ evidence trail properly.
Enterprise buyers and procurement teams already demand provable, verifiable proof of how an AI decides — today, deadline or not. Your clients are already being asked the question a policy binder can’t answer: “When a buyer or a regulator says prove it, what do we hand them?” A binder is a promise. It is not proof.
And plenty of the Act is already live: Article 5 prohibited practices since February 2025, GPAI provider obligations since August 2025, and the Article 50 transparency and Article 4 AI-literacy duties on their original schedule. VAL is how your clients hand over provable proof now — instead of a PDF they’re asking someone to trust.
VAL is a notary, not a scanner. It proves the record you keep is real, complete, and untampered. We say plainly what it does not do — that honesty is what makes it hold up in front of a regulator.
Every decision your client’s AI takes is recorded and cryptographically signed — the automatic logging the Act requires (Article 12).
Generated, versioned, and tied to the record — Article 11 and Annex IV, not a document that drifts out of date.
One file an auditor or regulator can verify independently — no trust in you, and none in us, required.
VAL proves the record is complete against a scope signed in advance — and flags a silently missing step even when every entry present is perfect.
Each decision can be re-derived: replay the policy on the recorded inputs and get the exact same outcome. A dashboard stores what you claim happened; VAL lets an auditor reproduce it — a fabricated decision can’t survive the replay.
Everyone can say their AI is auditable.
Almost no one can prove it’s complete.
Signing proves a single action is genuine. It does not prove the trail is whole. VAL is the only layer that catches a quietly dropped step — the exact failure a hostile operator, or an honest mistake, would otherwise bury under a stack of flawless receipts. That is the line no competitor on your client’s desk can draw.
And you prove all of it
without showing the data.
Selective disclosure: reveal only the steps you choose and seal the rest as a hash — content withheld. The auditor still gets cryptographic proof the record is complete and the disclosed steps are genuine, while your clients’ SSNs, financials and trade secrets never leave the building. The incumbents can’t do this — their whole model is to ingest your data to store it. VAL proves facts about the data while the data stays sealed.
And at the summit:
prove a fact, reveal nothing.
Zero-knowledge compliance proofs. Prove a property of a decision — “a human reviewed it,” “the outcome was compliant,” “no prohibited attribute was used” — while revealing none of the decision, the applicant, or the amount. Not redaction; mathematics: the verifier learns the property is true and learns nothing else. A governance dashboard stores your data to show it; VAL proves the claim without ever exposing it. This is the thing no one else in AI governance has.
You already own the client relationship and the advice. VAL is the deliverable your advice has been missing — and building it yourself is a product team and a permanent regulatory-maintenance burden, not a feature.
Your name, your logo, your support address, your client’s language — on every screen and every evidence bundle. We stay invisible.
Buy wholesale, sell retail, keep the spread — then layer your own onboarding and advisory on top.
Each client gets an isolated signing key and ledger, hosted in-region, backed by a DPA. One client can never read another’s record.
Provisioning a client is configuration, not code. You’re selling on day one.
The market’s AI-governance platforms hide their prices and sit between €30k and €150k per year. VAL has a price that is publicly available, credible, and clearly below them — so your customer sees the value and you keep the margin.
The first partners take on the risk of a new name, so they get the reward. 45% channel margin, raised to 55% for the founding group, plus a first-year price at half the rate for its initial customers — in exchange for a logo and a success story, with its rate protected as the market matures. A limited number of partners. After that, the standard program applies.
Your client’s system emits its activity to VAL through a drop-in SDK — Python or TypeScript.
Each entry is signed and appended to that client’s isolated ledger — their key, their data, in-region.
On demand, a signed evidence bundle — localised, mapped to the Act, ready for an auditor.
Your brand, your invoice, your relationship. The client never sees us.
Before a single slide, a prospective partner runs the evidence themselves — offline, on their own machine. This is that run. A clean record verifies; a silently dropped entry is caught where every other tool would report “all clean.”
If your clients already come to you for compliance, VAL is the deliverable that turns your advice into something they can hold. It fits the firms whose word carries weight:
A notary, not a scanner. We tell your clients exactly what VAL does — and doesn’t.
VAL proves the actions you instrument and the scope you sign are real, complete, and untampered. It does not claim to discover behaviour a black box was never wired to emit. In a market where an overclaim can void the very defence it was meant to provide, that restraint isn’t a weakness — it’s why the evidence survives a challenge, and why a serious partner can put their name on it.
And we shrink that boundary further than anyone: VAL comprehends your client’s code and proves the wiring itself is complete — every decision point is instrumented to emit, or it’s named and the build fails. Not just “the receipts are a complete set,” but “every place the AI decides produces one.” A governance dashboard can’t audit its own coverage. VAL can.
Tell us about your firm and we’ll set up a 30-minute session where you run the evidence offline — no deck. If it convinces you, we move to the founding-partner terms. A limited cohort, closing 31 December 2026.
No obligation. We’ll reply personally — nothing is automated, and nothing is shared.