AI Debt Collection Platform Exposed 6.6 Million Accounts and 315,000 Call Recordings

Two security vulnerabilities in CreditNirvana, an India-based artificial intelligence debt collection platform owned by fintech company Perfios, exposed approximately 6.6 million accounts and 315,000 call recordings, according to findings published Nov. 18 by independent security researcher Eaton Zveare.

The vulnerabilities potentially allowed unauthorized access to sensitive borrower information, including credit reports, identification documents, collection communications, legal records, and payment information. The findings raise important questions about data security and access controls as financial institutions and receivables management companies increasingly adopt AI-powered collection systems.

Security Flaws Exposed Millions of Collection Records

According to Zveare’s investigation, the vulnerabilities affected two components of CreditNirvana’s technology: its AI-powered call center and its primary collections management platform.

The first vulnerability involved a hardcoded application programming interface (API) key embedded in the call center’s software. The flaw could have allowed an unauthenticated attacker to access approximately 315,000 recorded collection calls, transcripts, and AI-generated recommendations.

Researchers also identified the potential to manage AI agents and create automated collection campaigns through the compromised system.

The second vulnerability, classified as critical, involved improper authentication controls that could have allowed an attacker to impersonate users, including administrators, and access sensitive information.

Zveare identified approximately 6.6 million potentially exposed accounts across three Indian financial services companies: Aditya Birla Capital Finance, BharatPe and International Asset Reconstruction Company.

Accessible records included credit reports, identification documents, payment receipts, settlement information, legal notices, and millions of collection-related communications.

The findings demonstrate how vulnerabilities in centralized collection platforms can potentially expose information across multiple financial institutions and servicing operations.

Vulnerabilities Patched Following Researcher’s Disclosure

India’s Computer Emergency Response Team (CERT-In) confirmed both vulnerabilities in an official security advisory, identifying them as:

  • CVE-2026-110474: A high-severity vulnerability affecting the CreditNirvana Agentic AI Call Center.
  • CVE-2026-110475: A critical-severity authentication vulnerability affecting the CreditNirvana Collections Platform.

Zveare reported the vulnerabilities to CERT-In on May 30, 2026. CreditNirvana corrected both issues on June 2, according to the researcher’s disclosure.

The vulnerabilities were formally documented on Oct. 26, with the broader findings published Nov. 18.

CERT-In recommends upgrading the call center to version 2.0.0 and the collections platform to version 4.8.2.

Importantly, the investigation established that information was accessible through the vulnerabilities. It did not establish that malicious attackers had exploited the flaws or stolen all the potentially exposed records.

CreditNirvana and Perfios did not respond to requests for comment from WIRED, which reported the findings Nov. 18.

What the Exposure Means for Debt Collection and AI Security

The incident highlights several cybersecurity considerations for receivables management organizations adopting AI-powered collection technology.

Collection platforms routinely process personally identifiable information, financial histories, recordings, payment records, and communications that may be subject to contractual and regulatory protection requirements.

AI-enabled systems can introduce additional security considerations when they connect customer information, communication tools, decision-making systems, and automated workflows.

The CreditNirvana vulnerabilities demonstrate why organizations should evaluate authentication controls, administrative permissions, API security, and vendor access to borrower information.

For collection agencies, creditors and debt buyers, vendor oversight should also address how automated systems store recordings and transcripts, restrict access across different clients, and prevent unauthorized changes to collection activities.

Organizations using third-party AI collection platforms may need to review security testing procedures, incident notification requirements, data retention practices, and access monitoring as part of their vendor management programs.

Although the identified vulnerabilities were patched months before the public research release, the investigation illustrates the potential consequences when collection technology combines large volumes of sensitive consumer information with insufficient access controls.

Published On: November 18th, 2026|By |Categories: Industry News & Announcements|Tags: |

Related Posts