ValidMind

As with others in this pocket the product governs artificial intelligence rather than obviously being driven by it, and the removal test would leave a documentation system, validation workflow and model inventory. What supports the grade is that the automated work is model work: generating validation documentation and test artefacts across model types including generative and agentic systems cannot be templated, monitoring capabilities produce real-time insight and alerts rather than scheduled reports, and the credit bureau partnership produced a distinct assistant product for this purpose. Held at B because no accuracy or capability detail is published for the automation itself.

The governance model for autonomous systems is stated more precisely than anywhere else in this index, reduced to three questions the platform answers about any agent: who approved it, under what conditions, and at what risk tier. Enforcement is described as real-time policy application rather than periodic review, which matches how agents actually operate. The platform sits within an institution's existing risk framework rather than replacing it. Held at B because nothing describes what the platform itself does automatically versus what a validator must attest to, which is the same question its product exists to answer for others.

The product is the discipline and the mapping is unusually complete, covering documentation automation, configurable validation workflows, model inventory, ongoing monitoring and outcomes analysis, aligned to named supervisory structures and verified as such by an independent comparison. The quantified problem statement is credible and sourced from customers, with roughly 30 percent of model risk team effort spent keeping records updated manually.

Held at B because no measurement is published for the platform's own work: no accuracy for generated documentation, no error rate, and no evidence of how often automated validation output requires correction before a validator will sign it.

A major credit bureau integrated the platform into its analytics environment in February 2025 so that banks can automate credit and fraud model documentation against several regimes simultaneously, and that firm's software division president is quoted publicly, which an independent comparison describes as a named enterprise proof point going beyond vendor self-description. A Canadian bank's chief risk officer is also quoted by name.

A global services firm entered a strategic partnership in 2025, and the company hired a chief revenue officer and a business development lead from an established analytics vendor. Backing totals 11.1 million dollars from a hedge fund's venture arm, an insurer's venture arm and a prominent AI fund.

No boundary statement was located. A platform validating models across many institutions accumulates knowledge of which model types fail validation, what examiners challenge and how findings resolve, which is precisely what would improve the product, and nothing states whether validation outcomes or documentation from one bank inform the platform's behaviour at another. The company's stated ambition to become a certifying authority offering validation on demand makes that question more consequential rather than less.

No data protection agreement, retention schedule, subprocessor list or deletion commitment was located. The platform holds model documentation, validation findings and inventory across an institution's entire estate, which collectively describes where a bank's decisioning is weakest, and where it reaches banks through a bureau's environment the data path involves a third party whose handling terms are equally undescribed.

No attestation, certification, trust centre or enumerated framework was located. A major bureau embedded the platform into its own product and a bank's chief risk officer endorses it publicly, so security assessment has been passed at a demanding standard, and nothing is published for other institutions to examine before beginning their own review of a system holding a consolidated account of weaknesses across their model estate.

This is the deepest regulatory mapping in the index. Five named regimes across four jurisdictions are addressed specifically: the original United States interagency guidance and its 2026 replacement issued jointly by the Federal Reserve, deposit insurer and comptroller, the United Kingdom prudential regulator's model risk principles, the Canadian supervisor's enterprise model lifecycle requirements, and European AI legislation.

An independent comparison confirms the mapping is structural rather than nominal, matching the three lines of defence construction, the United Kingdom's five principles and the Canadian lifecycle requirements. The company also publishes position papers on the transition between the old and new United States guidance, which is analysis rather than marketing.

The same irony applies here as at the other vendor in this pocket: a company whose product is governance publishes nothing about the governance of its own models. Bias testing, fairness evaluation and explainability appear as things customers must evidence rather than as capabilities described or as practices the company follows internally.

Given that fair lending analysis is among the principal reasons a credit model gets validated, and that the platform now documents credit models at scale through a bureau integration, the absence of any stated fairness capability is a notable gap.

No guarantee, indemnity or correction process was located. Responsibility for model risk sits with the institution under every regime the platform maps to, regardless of tooling, which is precisely why an explicit statement of what the vendor stands behind would matter. The stated ambition to become a certifying authority providing validation on demand would move the company toward assuming that responsibility, and nothing describes what it would then warrant.

No base model, provider, hosting arrangement or subprocessor is identified for the automation or the assistant products. The omission is pointed given the subject matter, since the supervisory guidance the company maps to treats undisclosed third party model methodology as a recurring examination failing, and a governance platform that does not disclose the provenance of its own components sits awkwardly against the standard it helps enforce.

The bureau integration is the substantive one, placing the platform inside an analytics environment banks already use for credit and fraud modelling so documentation is generated where models are built rather than afterwards, which addresses the practical failure of governance arriving late. Public developer documentation exists. The platform positions itself as a unified system of record across all models and AI systems in an enterprise. No named data science platform, deployment pipeline or model serving system appears, which for a model inventory is the connection that determines whether the record stays current.

No hosting provider, region selection, residency commitment or private deployment option was located. Banks in four jurisdictions are addressed, several with explicit expectations about where supervisory documentation is held, and model validation records are material a regulator may request directly, which makes processing location a question institutional buyers raise early.

No pricing, packaging or basis of charge was located. Model risk platforms typically price against inventory size or validation volume, and nothing indicates which applies, nor how the arrangement works where the platform reaches banks through a credit bureau's environment rather than directly.

Coverage follows the regulatory map rather than a customer list, spanning banking organisations subject to United States, United Kingdom, Canadian and European requirements, with a named deployment at a Canadian bank and distribution through a bureau serving lenders internationally. The platform is described as covering all models, AI systems and record types across an enterprise rather than one model class. Held at B because institution size bands are not described and the buyer remains banking specifically rather than the wider set of regulated firms carrying model risk obligations.