Four properties. One receipt.
A verification layer earns its place when someone outside the building can check the work.
See the standard →
Receipt · NEXI-VS
Actoragent://acme-ops/dispatch-7
Actionmodel.infer · settlement.approve
Inputssha256:2b9c4d8a…f107
Resultsha256:9f3a1c7e…b204
Re-derived✓offline · stdlib verifier
✓ Verified
01
SignedIN BUILDed25519 signing, with production key management and a certificate chain — hardened inside a Verify Standing.
02
Re-derivableLIVEAnyone holding the receipt re-computes the result offline, from the bytes it ships — no NEXI infrastructure required.
03
AnchoredIN BUILDA live OpenTimestamps anchor-runner exists — it stamps receipts to public Bitcoin calendars (first soaked 2026-05-22); demo bundles aren’t anchored yet. Production daily anchoring hardens inside a Verify Standing.
04
ConformantLIVEOur releases are Sigstore-signed against an open spec today; your system is conformance-checked against that spec as the slots are wired in.
Run a verification.
Receipt · NEXI-VS · draft
Actor
—
Action
—
Inputs
—
Result
—
Signed
—
Anchored
—
Pending
Verification waterfall
5 checks
Interactive demo on illustrative data, run in your browser — not a real audit. The SHA-256 re-derivation and ed25519 signature verification are real cryptography computed right here; public anchoring, claim grounding and citation checks are simulated in this demo. (In the product, byte-offset span attribution and citation-class checks are built; a live OpenTimestamps anchor-runner exists; semantic entailment is not yet built.) Verify flags problems. It does not rubber-stamp.
Action
Checks
Proof checks
Verification feed
Run a verification to stream the feed.
Checks run0
Checks passed0
Receipt statusPending
Re-derivableyes
Built to become the standard the AI era runs on.
A standard is not declared. It is adopted — and the first major adopter in a vertical defines what trustworthy AI looks like there.
01
Open, not ownedThe methodology and verifier are public — no private benchmark, no black box.
02
Adoption sets itEach conformant system makes the next one easier for a buyer to ask for.
03
One bar, everywhereEvery NEXI intelligence ships with the same receipt — one mark of proof.
Verify enters one vertical at a time.
01
Verify · Energy
Dispatch, trading and emissions reporting — every AI decision a regulator or counterparty has to trust.
02
Verify · Life Sciences
Drug discovery and trials, where regulators already ask sponsors for credible model output.
03
Verify · your vertical
Reference pilots span more than twenty domains and data formats today — from healthcare and legal to climate, finance and life sciences — backed by a 46-class adversarial corpus. Your vertical edition is co-built over the engagement.
Two doors. One has a price, one has an introduction.
The Verification Audit is one fixed-scope verification at a public price. A Verify Standing is the operated, recurring relationship — by introduction, priced to you, never to a list.
The offersTwo doors, plus enterprise
01
Verification Audit
A fixed-scope verification of one AI system, re-derived end to end from its raw inputs. Public checkout, delivered in two weeks.
≈ $25KCheckout →
02
Verify StandingSTANDING
The operated, recurring layer — an annual platform floor, a metered per-verification line, and one outcome line agreed up front. Entered by introduction; the Audit credits in full toward year one.
By introduction
03
Enterprise
Multi-system, multi-jurisdiction or sovereign scope — conformance terms scoped to you.
Request a quote
What a Standing holdsHeld for Principals
01
The held rateA recurring annual rate on defined scope, held for a bounded term, then a bounded step. Earlier accounts hold the lower rate.
02
A voice in the methodologyDirect input to the contract set and the rigor ladder, with the people building Verify. The Principal is consulted; NEXI decides. Influence, never authorship.
03
The mark and a referenceA named, mutually approved reference, and the right to carry the NEXI verification mark.
04
First access, namedFirst access to new contracts and attested builds as they ship, and a Principal designation, named with your consent.
Why a Standing recursThe ongoing value
01
The proof is permanent. Its currency is not.A verification is a cryptographic proof of what was checked, and when — that never decays. Its binding to the live model set, threat model and regulatory frame does.
02
Models and attacks keep movingA verifier hardened against today’s attack classes is not hardened against tomorrow’s. A Standing keeps your conformance corpus current and your verifier builds attested as that work hardens.
03
Independence you cannot self-hostA verifier you run yourself is not an independent one. A Standing operates the public anchor, the transparency log and the managed re-verification a regulator or counterparty can check.
The fine printRead before you sign
01
Build stateVerify is in build. Re-derivation, the rigor model and daily public anchoring run today. Attested builds, signed DSSE envelopes and graph-wide revocation are in development.
02
Open and paidThe standard and the verifier are published open and source-available. The annual fee is the operated layer — the current adversarial corpus, attestation, anchoring and re-verification.
03
The standardNEXI authors the specification and contributes it to an open standards body. NEXI does not sit in judgment of conformance to its own standard.
04
The rateThe Audit price is fixed and public. The Standing rate is quoted per account against modeled value, held for a bounded term, and never published. Earlier accounts hold the lower rate.
05
What the proof coversVerification is a necessary part of accountability, not the whole of it. Verify supplies the evidence a review needs; the judgment stays with you.
Rates follow NEXI’s Verify Value Model — modeled value, a pricing estimate, never a discount. The standard and verifier are published open; the operated layer stays NEXI’s. Invoicing: billing@nexiverify.com.
Common questions about AI verification.
01
What is NEXI Verify?NEXI Verify is the verification layer for AI. Every action an AI system takes becomes a re-derivable receipt — a record anyone can independently re-check, offline, without trusting the system that produced it.
02
What is AI verification?AI verification is producing independent, re-checkable evidence that an AI system’s output was computed correctly from its stated inputs — proof of the work, not a score you are asked to trust.
03
How is verification different from AI monitoring?Monitoring reports what an AI system says about itself. Verification re-derives the result from the raw inputs, so the answer holds even if the original system is wrong, offline, or gone.
04
Who needs AI output verification?Teams whose AI decisions face outside review — regulators, auditors, courts, counterparties — especially in energy, life sciences, financial markets and aerospace.