Strise

Relationships, not records, are the product: entities, directors, ownership chains and sanctions exposure live in one continuously updated graph, with natural language processing and false positive suppression around it. Without the models, what remains is licensed registry data, another company's product.

Each review opens with an AI risk summary that the analyst can edit, along with recommended actions, before submitting, and AI adverse media search sits inside the standard screening review. Routine cases are routed to automation and complex ones to deeper review, but no threshold says which may close without a person.

Operational outcomes are published, but precision, recall and a validation pack are not, though European banks are expected to validate the financial crime models they rely on. Marketing has claimed fewer false positives than any other screening tool without naming a benchmark or evaluator.

Endre Jo Reite of BN Bank reports false positives down 30 percent, and executives from BDO and British Land are quoted by name without figures. Platform figures, 80 percent of routine cases automated and complex cases cut from three hours to 30 minutes, attach to no customer.

Customer data reaches third party providers only under negotiated contracts, and results are not shared across tenants. Whether documents, ownership corrections or review outcomes train the models goes unaddressed, though customers edit ownership maps that others share.

Auth0, Intercom and Mailgun appear with their locations on a public subprocessor list, and individuals have rights to access, modify and delete their data. No retention schedule or deletion timeline is published.

ISO/IEC 27001:2022 and a SOC 2 Type 2 attestation by Prescient Assurance are named, with a trust center on its own subdomain and a public status page that makes availability observable. The trust center runs on Vanta, and its documents do not open as public pages, so how reports are released is not clear.

Outside the regulatory perimeter, Strise states no license, registration or sandbox participation, leaving the anti money laundering obligations with its customers.

Names in non Latin scripts, or with several accepted romanizations, draw more sanctions and politically exposed person flags simply for how they are written. Fairness evaluation and a breakdown by name type or jurisdiction are unpublished, and the consequence lands on the person or business at onboarding.

A business wrongly matched to a sanctions record, or scored high risk on a mistaken ownership resolution, may be declined without learning a model produced the finding. Strise publishes no commitment behind the output, and nothing says whether a correction reaches other institutions on the graph.

The same subprocessor list carries no external model provider. It dates from April 2025, before the AI summary and adverse media features, and does not say which models run them.

Kyckr supplies live data from more than 300 registries, Signicat joins it in a no code identity offering, and Stacc embeds Strise in Nordic lending software, alongside a public API. Systems of record on the buyer's side, from core banking to CRM, are not named as direct integrations.

All customer data is processed and stored on Google Cloud servers in Belgium. On premises deployment and region choice are not described.

Licensing is quoted per customer, with no pricing, tiers or unit published. For a platform sold on cost reduction, how price follows portfolio size, users or screened entities is unknown.

Named customers span banking (Nordea, Danske Bank), insurance and pensions (Fremtind, KLP) and audit and advisory (PwC Norway, EY UK), among other segments. Roughly 70 percent of top tier Nordic banks use it, by its account, and the base remains concentrated in the Nordics.