Skip to content
Company

A trust assessment for any entity — ultimately citable, like a rating.

GriffynX Intelligence Labs builds trust infrastructure. The standard we hold ourselves to is that other institutions' systems make decisions on our assessments, without a person in the loop, and remain willing to be held accountable for those decisions. Every choice in the product is checked against it.

Our position

Trustworthiness is computed from conduct, footprint, network and history.

It is neither creditworthiness nor identity verification; those are inputs, not the product. The output is always an assessment: a score, a computed confidence interval, the evidence chain, provenance for every fact, and the methodology version. A label — decline, review, accept — is the institution's policy applied to our assessment. It is never ours.

That distinction is the whole business. A vendor that supplies the verdict assumes none of the accountability and permits none of the explanation. We supply the evidence and the calibrated uncertainty, and the institution retains both.

What must hold, simultaneously

  1. 01

    One evidence graph per entity — across sources and across time, each fact carrying its provenance, in one place that reasons over the whole of it.

  2. 02

    Assessments computed and defensible — a score with an interval derived from weighed evidence, with the chain attached, so the reason for a decision is always available and always accurate.

  3. 03

    Policy belonging to the customer — we supply evidence and calibrated uncertainty; we never supply the verdict, and we do not accept the liability that would come with it.

  4. 04

    Memory that compounds — every assessment, closed case and confirmed outcome improves the next answer, within a tenancy by default and across tenancies only under a legal model.

  5. 05

    Consent, purpose and jurisdiction built in — every call carries a purpose; every datum a class and a retention regime; every jurisdiction is a pack of data rather than a branch in code.

Where we begin

India, lending and payments. A pack and a policy, not an identity.

Highest incidence, clearest integration point, best outcome data. India's authorities shared 3.2 million first-layer mule accounts with its banks; of everything reported lost to cyber-enabled fraud since the national portal opened, a fraction of one percent has been returned. The reasoning core is identical in Mumbai, Manila and Munich. Only the packs differ.

Depth on rails nobody global has

UPI, +91 numbering, DLT-registered SMS headers, national cyber-crime reporting and vernacular fraud scripts — first-class artifact types, not a generic string field with an annotation.

Regulation is the tailwind

The RBI's FREE-AI framework names explainability, accountability and fairness as guiding principles. The UK reimburses victims of authorised fraud, Singapore allocates liability, Australia legislates penalties. An unexplainable assessment is now a cost.

A consented entrance

Discover has individuals and entities voluntarily associating themselves with identifiers, with a stated purpose. No scraped-data vendor has that, and it cannot be retrofitted.

Global by construction

Nothing below the pack boundary contains a country, currency, regulator or identifier format. Entering a market is configuration, not a release.

What we decline to build

The list that keeps the product honest.

Each of these would win revenue in the short term and destroy the property the business depends on. They are refused as a matter of standing policy, not case by case.

A blacklistA rule engineAn unexplainable scoreA conversational wrapper over a language modelCyber threat intelligenceBespoke per-customer deploymentsBiometric inferenceAny confidence a person or a model can set by hand

We also decline to infer protected or proxy characteristics, to supply the verdict on an institution's behalf, and to open an investigative case without an authorisation on the record. The last of these has cost us conversations. It is not negotiable.

Where the platform stands

Two surfaces in production. Two on the roadmap, contracts already published.

Every surface returns the same assessment from the same kernel. What is served today is marked as served; what is designed and published as a fixed contract is marked as that, here and in the documentation.

Live

Discover

The evaluation surface at discover.griffynx.com. Structured intake, the written assessment, and the verdict with its basis.

Open
Live · v1

Trust API

Inquiries served; forty-three endpoints published as a fixed contract at api.griffynx.com. X-Ray evaluations run on it.

Open
Roadmap

Investigate

Analysts, cases and evidentiary bundles. The contract — cases, members, intake, expansion, hold, review, export — is already published and fixed.

Final

Rating

The published, citable form. It follows calibration measured per pack against reported outcomes, and not before.

Who we are

GriffynX Intelligence Labs Pvt Ltd

Founded by Kunal Gupte and built by a small team in India. We work directly with our design partners and answer every enquiry ourselves, ordinarily within the same working day.

For investors and partners

Counterparty trust is priced badly everywhere, and the market for deciding it is being rebuilt by regulation that now demands a reason for every adverse decision.

We are building the layer that supplies the reason: evidence carrying its provenance, an assessment that is computed rather than asserted and can be reproduced years later, and a ledger that improves with every decision made upon it. Both products are in production and can be evaluated without speaking to us first — which we regard as the strongest statement we can make about either.

Evaluate it before you speak to us.

Both products are available without an introduction: run a check on Discover, or read the contract your engineering team would integrate against. A conversation is only necessary for a pilot or a key.