Ripjar upgraded the adverse media risk classifiers in its ULTRA engine, which read each news article to decide whether it relates to a specific risk and which person is involved, moving them to a new generation of machine learning model. The same update improves entity matching so screening can tell apart parent companies, subsidiaries and similarly named group entities, and handles more complex naming conventions, including Arabic names. Both feed Screening Assistant, the agentic triage feature that closes low risk alerts on its own.
Our readA model change in the classifiers that decide adverse media risk is something model risk teams need to log and retest, and it changes which alerts the AI closes without an analyst. Ripjar named no model and published no accuracy figures, so buyers should ask for validation evidence on their own populations.