Ripjar

Name matching is stated at 94 percent better than Levenshtein distance, the string algorithm a screening system uses without a model, so the version without models is one Ripjar itself calls materially worse. Adverse media extraction across more than six billion articles and entity resolution across scripts are learned problems, and Screening Assistant makes close and escalate judgments.

Screening Assistant closes low risk alerts automatically, routes edge cases to a person with evidence and records every closure with its reasoning, credited with a 77 percent cut in human effort. A closed alert is an investigation that never happens, and no confidence threshold, configurable floor or required sampling of closures appears.

False positive reduction of up to 91 percent comes with no true positive retention figure, though fewer false positives can come from detecting less and the reduction is produced by automatic closure. Recall against a known adverse set, validation method and drift disclosure are unpublished.

Screened names at Dow Jones rose from under one million to over ten million, and more than 400 customer organizations include six of the 29 global systemically important banks. Review time falls from 20 minutes to 3 by Ripjar's account, and Nomura, UBS and Barclays appear among the logos, though both bank case studies are anonymized by tier.

Whether customer data, analyst decisions or closed alert outcomes improve the models is not stated. No acceptable error rate is published for agentic closure either.

Privacy, cookie and applicant notices are published, but nothing covers platform processing, retention for the entity profiles kept so decisions persist, or correction of a wrong adverse media link. Despite sales into the United States, the Gramm Leach Bliley Act (GLBA) goes unmentioned.

ISO 27001 and SOC 2 marks appear on the site, the right pair for its buyers. Certificate, scope statement and SOC 2 report type and period are all unpublished, and SOC 2 yields an attestation report rather than a compliance state.

Material engages the senior managers and certification regime and Singapore's accountability guidelines in substance, while Ripjar Ltd claims no financial services authorization. Its signals intelligence heritage implies public sector assurance, but no accreditation or clearance is named and no position on the EU AI Act is stated.

Matching across more than 400 languages and 20 scripts fails differently by script and naming tradition, so the odds of a wrong sanctions match, or a wrong clearance, depend on where a name comes from. Automatic closure compounds that, a rejected link persists under the entity model, and no per script accuracy or fairness testing is published.

Every decision, automatic closures included, is time stamped, source linked and reconstructable, which helps a customer defend itself to an examiner. Ripjar publishes no terms, warranty or liability cap, so what it owes if Screening Assistant closes an alert that should have escalated is unknown.

ULTRA, the engine behind entity resolution and name matching, is an in house build, and customers supply their own data feeds, so the core chain is short. Summaries were announced in 2023 to use large language model technology, yet no model provider is named and nothing says whether screened subject data reaches one.

Customers bring their own sanctions lists, politically exposed person data and any adverse media supplier under a data agnostic design, and Dow Jones is both investor and customer. API reference, connector catalog and developer portal are not published, so integration effort cannot be judged before contact.

Cloud deployments and on premises installations are both referenced, with cloud described as typically faster, which documents an on premises route. For the hosted service, where entity profiles and decision history sit with Ripjar, no cloud provider, region or residency commitment is given.

A demonstration and a free 30 minute screening diagnostic are the entry routes, with no price, unit of billing or contract length published. The one cost figure on offer, $30 to $70 to review a false positive, is an industry estimate rather than Ripjar's price.

Global banks, corporates and regulators are named buyers, backed by tier one and tier two bank case studies. Credit unions, community banks and insurers get no material.