Salv

AI Powered Insights reviews a screening match and explains its reasoning without ever closing an alert, while screening, monitoring, risk scoring and intelligence sharing all run on rules and matching an institution configures. Remove the AI and the product does its job unchanged.

The AI is held out of the decision: AI Powered Insights explains a match and never closes an alert or decides. Automation that does close alerts is rule based and visible, since a clearance rule closes an alert only on conditions the institution sets and the rule stays shown against the alert, and no alert status can change without a note that cannot be deleted.

Accuracy figures for the AI review are not published. AI Powered Insights explains the reasoning behind each match assessment, customer risk scores come from weighted factors and level ranges an institution sets itself, and scenario versions cannot be edited after the fact, with configuration changes logged.

LHV Bank in Estonia cut the share of transactions reviewed by hand from 1 percent to 0.25 percent with Salv screening, so one agent handles more than 13,000 alerts a month, with median and average resolution times of 16 and 45 seconds. On Salv Bridge, LHV releases 40 percent of payments automatically at any hour and processes requests for information 2.5 times faster. These figures measure the screening workflow and the sharing network, not AI Powered Insights, which carries no figure.

Bigbank is named on its move into everyday banking, now live in nine countries, and an unnamed Nordic bank raised successful authorized push payment fraud reclaims from 10 to 80 percent with Salv Bridge. Platform figures, more than 130 institutions and around 96 million transactions screened a month, attach to no single customer.

Whether client data or alert decisions train the AI Powered Insights model, and how one institution's data is kept from another's model, is not stated. Salv Bridge shares fraud and sanctions intelligence between institutions by design, under the cooperation rules the network is built around.

A subprocessor list is not set out, but the rest is specific. Personal data processed for a client sits in Frankfurt, clients own their data and can export it through the application programming interface (API) at any time, and on exit Salv returns or deletes it on request or within one month of the end of service, with breach notice inside 48 hours and a named data protection officer contact.

Neither document can be downloaded; both are shared during a request for proposal. Salv's SOC 2 Type 2 report covers the Security category, follows International Standard on Assurance Engagements (ISAE) 3000 and is issued by Thoropass Assurance, and its ISO/IEC 27001:2022 certification comes from KPMG IT Certification, both renewed annually, with annual external penetration testing.

Salv contracts as an information and communication technology third party provider under the EU Digital Operational Resilience Act, giving clients, their auditors and competent authorities rights of access, inspection and audit. Its intelligence sharing is framed against Article 83a of the Payment Services Regulation and Article 75 of the Anti Money Laundering Regulation.

Fairness testing of name matching or of AI Powered Insights is not published, though fuzzy name matching across transliterated Cyrillic, Greek and Chinese entries is where unequal false positive rates would show. The governance answer Salv gives is structural: the AI is informational only and a person decides.

A guarantee or indemnity behind a screening or monitoring outcome is absent. What a supervisor or affected party can rely on is the audit trail: every change is timestamped with what changed and by whom, notes cannot be deleted, cleared alerts stay on file with the rule that cleared them, and alerts export in an audit template.

The model behind AI Powered Insights goes unnamed. The rest of the chain is set out: Dow Jones supplies the sanctions, politically exposed person and adverse media lists, and Amazon Web Services hosts client personal data in Frankfurt.

Core banking and case management products on the institution side are not named. Salv publishes documentation for its application programming interface (API): person, transaction, screening, monitoring, risk and alert endpoints under OAuth2 client credentials, webhooks that push new alerts and risk level changes in real time, a sandbox environment and a CSV upload that loads data in batches of 1,000 rows. Dow Jones supplies the screening lists.

Private cloud and on premises deployments are available alongside the hosted service, which runs on Amazon Web Services with personal data stored across all three availability zones in Frankfurt, Germany.

Price, tiers and the unit of charge are unpublished, and the route in is a conversation with Salv. Security questionnaires and evidence packs also travel through that sales and request for proposal route.

Banks, fintechs and payment service providers are the buyers, with dedicated material for intelligence sharing obligations under the EU Payment Services Regulation and the Anti Money Laundering Regulation. Material is organized by product and use case rather than by institution type.