cd /news/ai-safety/an-ai-debt-collection-service-expose… · home › topics › ai-safety › article
[ARTICLE · art-172034] src=wired.com ↗ pub= topic=ai-safety verified=true sentiment=↓ negative

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

Two vulnerabilities in CreditNirvana, the India-based AI debt collection platform owned by fintech company Perfios, exposed 6.6 million accounts and about 315,000 call recordings from its agentic call center, according to research published Wednesday by independent researcher Eaton Zveare. India's Computer Emergency Response Team (CERT-In) rated one flaw, CVE-2026-110474, high severity and the other, CVE-2026-110475, critical; the bugs were patched in June and disclosed publicly in October. Zveare told WIRED the high-severity flaw also could have let an attacker manage AI agents and direct their own agentic campaigns, while the critical flaw exposed credit reports, ID cards, court records and legal documents tied to three prominent Indian financial services customers.

recommendation

Audit agentic platforms for unauthenticated access to agent management and call-center data; the CreditNirvana flaws let attackers impersonate admins and direct AI agents before a one-day fix.

more in playbook: Securing agents and AI tools
by read2 min views4 publishedNov 18, 2026
An AI Debt Collection Service Exposed 6.6 Million Accounts and 315,000 Call Recordings
Image: Wired AI

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. Owned by the Indian fintech company Perfios, the platform is built to “read a borrower’s full context, decide the next best action, and execute it autonomously, continuously, and within policy.”

The bugs were patched in CreditNirvana's systems in June and were disclosed publicly in October. But independent researcher Eaton Zveare, who found and disclosed them, notes that, while the vulnerabilities were active, they exposed personal information on millions of borrowers. Details about financial collections are inherently—perhaps obviously—sensitive, because they relate to unpaid accounts or debts and often involve extensive documentation of people's personal histories.

“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. There’s just so much in there that makes it a very sensitive issue.”

CreditNirvana and Perfios did not return WIRED's requests for comment. Zveare says he worked directly with India’s Computer Emergency Response Team (CERT-In) to get the vulnerabilities fixed. CERT-In categorized one vulnerability as “high” severity and the other as “critical” severity.

The high-severity vulnerability impacted the call center data, exposing hundreds of thousands of calls as well as recommendations for next steps generated by the platform after calls were complete. It could also have been exploited to allow an attacker to manage AI agents within the platform and direct their own agentic campaigns.

The critical severity vulnerability impacted CreditNirvana's main platform, and Zveare says it specifically exposed the data of three corporate customers—all prominent Indian financial services firms—using CreditNirvana for their collections. The data also exposed millions of individuals’ accounts, which contain credit reports, ID cards, case management details, and legal documents, along with certain types of communications, including millions of emails, texts, and call transcripts related to financial services, unpaid accounts, and collections. The vulnerability also could have been exploited by an attacker to track the location of human debt collection agents and for some administrative privileges within the system.

“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.

Zveare says he is relieved that the issue was resolved quickly, but is conscious that the bugs were potentially present in CreditNirvana's systems for years. As AI is integrated into more and more automated systems, synthesizing data from a greater number of sources and using it to draw conclusions, Zveare says, the episode serves as “an important example of an AI-powered service having a very sensitive data exposure.”

── more in #ai-safety 4 stories · sorted by recency
── more on @creditnirvana 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/an-ai-debt-collectio…] indexed:0 read:2min 2026-11-18 · —