Two vulnerabilities in the India-based AI-driven debt collection platform CreditNirvana exposed 6.6 million accounts and about 315,000 call recordings from the company's agentic call center, according to research released on Wednesday. The platform, owned by the Indian fintech company Perfios, is designed to read a borrower's full context, decide the next best action, and execute it autonomously, continuously, and within policy.

Deploying autonomous agents into debt collection shifts operational exposure into high-severity infrastructure flaws that jeopardize customer data. As independent researcher Eaton Zveare explained, some of the largest fintech firms across India use CreditNirvana to autonomously collect debts, meaning the vendor concentrates credit scores, call recordings, transcripts, court records, and other sensitive documents across millions of borrowers.

"They fixed it in one day, which was great, but this is really a pretty significant case," Zveare tells WIRED. "Some of the largest fintech firms across India use this company to autonomously collect debts, so they naturally hold data, including credit scores, call recordings, transcripts, court records, and other sensitive documents."

CreditNirvana and Perfios did not return WIRED requests for comment regarding the incident. What operational exposure did these flaws create across the platform's customer base?

Agentic Hijacking and Enterprise Records

The security impact split across two distinct layers of the software architecture. Eaton Zveare worked directly with India's Computer Emergency Response Team, known as CERT-In, to address the flaws, with the agency categorizing one bug as high severity and the other as critical severity. The high-severity vulnerability compromised the call center environment, exposing hundreds of thousands of calls along with platform-generated recommendations for subsequent actions. The flaw created an attack vector allowing an adversary to manage AI agents within the platform and run their own campaigns.

The breach went beyond conversational data into core financial operations. The critical severity flaw hit CreditNirvana's main platform, exposing the records of three corporate customers that operate as prominent Indian financial services firms using the service for debt recovery. This critical bug opened access to millions of individual accounts containing credit reports, government identification cards, case management files, legal filings, and millions of text messages, emails, and call transcripts.

"Successful exploitation of this vulnerability could allow the attacker to enumerate platform users, establish sessions as arbitrary users including administrative users, access and modify sensitive information with the privileges of the impersonated user," CERT-In wrote.

Beyond data exposure, an attacker could exploit the critical flaw to track the physical locations of human collection agents and acquire administrative privileges inside the environment.

Long-Term Platform Exposure

The primary mitigating factor rests on rapid remediation: CreditNirvana patched the vulnerabilities in June, completing the fix within a single day before public disclosure in October. Yet that rapid turnaround does not eliminate the historical risk profile. Eaton Zveare pointed out that these vulnerabilities were potentially present in CreditNirvana's systems for years before discovery.

A system engineered to enforce financial compliance autonomously ran for years with doors open to administrative hijacking and mass data exfiltration.

For comparison

An AI-driven platform built to enforce strict debt-collection rules ended up leaving its core infrastructure wide open for years. India-based CreditNirvana accumulated sensitive borrower records from major financial institutions, yet architectural flaws let outsiders manipulate internal agents and launch unauthorized calling campaigns. While engineers patched both flaws within a day in June, swift bug fixing cannot erase the years these backdoors stayed exposed in the codebase. A system designed to automate financial compliance ultimately left 6.6 million borrower records exposed.

Artificial IntelligenceAI AgentsAI in FinanceCybersecurityEnterprise AI APIs