Vulnerability / research

Researcher accesses AI debt-collection systems through missing authentication checks

Reported

Published here

What happened

Eaton Zveare demonstrated access to CreditNirvana’s Agentic AI Call Centre through a client-side API key and to its Collections Platform through unchecked authentication. His November 18 account describes access to 315,000 call recordings and 6.6 million accounts across three client deployments.

Parties and reporting

Where this happened

Victim location not disclosed. Real organizations or systems were affected or targeted, but the sources do not say where. Explore the world map.

AI tools and systems: CreditNirvana Agentic AI Call Centre, CreditNirvana Collections Platform / CollectionWorkz, AI call transcripts and summaries

Outcome

Zveare reports that both flaws were fixed on June 2 after coordinated disclosure. CERT-In identifies Call Centre 2.0.0 and Collections Platform 4.8.2 as corrected versions. Neither source establishes malicious exploitation, bulk copying of all records or completed customer notification.

Operating lesson

Validate authentication and authorization at every API, remove shared credentials from client code and verify tenant separation.

Evidence and limits

Accounts and recordings are exposed scope, not counts of distinct people or confirmed theft. The authentication failures belong to CreditNirvana’s integration, not an asserted defect in Keycloak or Amazon Cognito. October 26 is the earliest verified public advisory; November 18 is later detailed coverage. Remediation was not independently retested.

Sources