India's cyber agency says an AI-assisted gang is still draining banks and fintechs through their own payment APIs, and re-hit firms that had patched

India's national cyber agency, CERT-In, has put out a critical-severity warning to banks, lenders, payment firms and crypto exchanges: an AI-assisted group is moving money out of them by calling their own payment systems, and it hasn't stopped.

What happened

On Wednesday 4 November, CERT-In issued advisory CIAD-2026-0058. It updates a warning from 9 October about attackers abusing payment and disbursement APIs to make unauthorised fund transfers. The finance-sector response team inside CERT-In has now named the group behind it 'CRAFTY FALCON'.

CERT-In describes it as an AI-assisted threat actor that surveys its targets with machine intelligence, works at machine speed, and strikes by invoking the victim's own payment and disbursement APIs. It says analysis of incidents reported by non-bank lenders, digital lending platforms, fintechs, payment aggregators, payment service providers and crypto exchanges shows a significant escalation.

The AI part

According to the advisory, the group uses several publicly available AI-driven offensive security tools, including the ARTEX family, the PentAGI multi-agent penetration testing system, the Cairn attack search engine and the CyberStrikeAI framework. CERT-In says these tools run complete attack chains on their own, churn out and refine large numbers of tailored attack scripts quickly, and adapt their payloads to how the target responds.

How the money moved

Reported tactics read like a checklist of soft spots. The attackers went in through non-production environments such as test sites, then moved into live production using credentials, tokens or network paths shared between the two. They abused cloud and serverless setups to run commands, reverse-engineered apps to build tailored exploits, and sent API calls from modified copies of the apps, getting around controls that assumed only the real app would ever call the API. They routed traffic through trusted third-party services to look legitimate.

Before moving bigger sums they sent test transfers of ₹1. Payment aggregators' merchant settlement APIs were abused to move funds, and crypto exchanges were hit with stolen wallet keys.

Patched, then hit again

The worst line in the advisory: CERT-In says it is still receiving reports of successful attacks, including from firms that had already received the earlier restricted advisories and done remediation. Entities were deliberately re-targeted after incomplete fixes, and attacks were timed for holidays, weekends and other low-monitoring windows. CERT-In hasn't published a total loss figure.

Its advice includes round-the-clock monitoring with more staff and lower alert thresholds on holidays, strict separation of test and production, no shared credentials or API keys, server-side checks on every transaction, signed API requests with nonces and timestamps, and app attestation to spot tampered apps.

Lesson for builders

Your API is the bank vault now. If a test environment shares a key with production, or your server trusts that only your own app will call it, assume an automated attacker will find that within hours. Validate every money-moving request on the server, and treat a ₹1 transfer from a strange source as an alarm, not a rounding error.