CleverChain

Data comes from third party providers, reconciled by CleverData, so the value sits in what the VERA agent and its companions do with it: linking entities, reasoning across ownership and control, and reaching conclusions calibrated to each institution's policies. Remove the models and what remains is access to other companies' data.

Immaterial changes become silent timestamped updates, material issues become internal alerts with evidence, and some trigger client outreach, with control checklists mapping each item to evidence for supervisory review. Who signs a VERA conclusion before it enters a compliance file, and whether automated outcomes are sampled, CleverChain does not say.

Each conclusion carries its reasoning and the source document behind every finding, shell company assessments present the case against the finding where exculpatory evidence exists, and model versions, sources and outputs are logged. Accuracy and false positive measures, a validation method and any mapping to supervisory model risk guidance are missing.

Customers are described by tier rather than by name, such as a top five global banking group and international private banks, so no outcome is tied to a named institution. Coverage of 5.6 billion entities and more than 350 registries measures data reach, and 2026 awards from Chartis and RegTech Insight measure standing rather than results.

Agents run on third party language models under zero data retention agreements, CleverChain states it does no proprietary model training, and model selection is configurable per tenant. Whether findings assembled for one institution enrich the entity data another sees is a separate question the commitment leaves open.

The privacy policy covers website visitors, job applicants and service users, with retention set for account and recruitment data and UK and EU GDPR rights named. People who appear in its investigations and risk graphs, and the lawful basis for processing open source material about them, are outside its scope, which matters for a product that assembles identified individuals and their relationships into a continuously updating graph.

ISO 27001 certification is held and SOC 2 is in progress, by CleverChain's own account, with encryption, audits and staff training described in the privacy policy. No trust center or retrievable certificate is published, leaving scope and the SOC 2 timeline unstated.

Selection for the FCA's regulatory sandbox puts the due diligence capability through live testing under supervisory oversight. CleverChain has also signed the EU code of practice on transparency of AI generated content, though the sandbox tests the product and the compliance decision stays with the buyer.

Bias is one of the measures in the sandbox test, alongside transparency and explainability, which matters where name matching and open source material disadvantage non Latin names, common surnames and businesses in thinly registered jurisdictions. The method, findings and any error analysis by jurisdiction or name form are not published.

Warranties, service levels and correction obligations are absent from what CleverChain publishes, though its traceable conclusions let an institution challenge a finding before acting. A business denied banking on an adverse conclusion has no described route back to the party that produced it.

Named data providers include Experian, LexisNexis and Kyckr, and the model layer is described as third party language models chosen per task. Which providers supply those models is never stated, nor is a subprocessor list published.

A REST interface reaches onboarding, compliance and case management systems, Model Context Protocol access serves customers' own agents, and integration is said to take under a week, with TransactionLink as a named onboarding partner. API documentation is not public, and no core banking or case management vendor is named.

Cloud native on AWS across multiple availability zones, the platform splits a control plane from isolated regional data planes and offers regional endpoints where data providers support them. Single tenant, private cloud and on premises options are not described.

Intelligence is billed on usage with an annual minimum commitment and pro rata overage, and policy automation sells under tiered annual licenses, enough for a buyer to model cost before contact. Rates, tier boundaries and the minimum itself stay unpublished.

Financial institutions from global banking groups to private banks are served alongside aerospace and defense organizations, law firms and risk consultancies. Material is organized by agent and function rather than by institution type.