India's national cyber security agency, CERT-In, on 20 November 2026 issued a critical-severity advisory from New Delhi warning that criminals are making unauthorised payments by tampering with business records at technology service providers, including firms that run government-linked payment processes. In the same advisory, numbered CIAD-2026-0062, the agency said its own threat hunting had found attackers using AI-driven "agentic offensive frameworks" against enterprises, government bodies and research institutions.

The Indian Computer Emergency Response Team, which works under the Ministry of Electronics and Information Technology, did not say how much money had been paid out wrongly, which organisations were hit, or who was behind the incidents.

What did CERT-In find in the payment fraud cases?

CERT-In found that attackers did not need to break into payment systems directly; instead they changed the records and approval relationships that decide who gets paid, and then let normal payment processes send the money. The agency said the reports it received involved organisations outside the banking and financial sector, among them service providers handling government-linked payment workflows and platforms that make vendor or partner payouts.

According to the advisory, the incidents involved misuse of business applications and workflows operated by service providers, the use of trusted systems and established approval steps so that transactions would not immediately look suspicious, and gaps in independent checks, reconciliation and review of changes to business records.

The agency's central warning is that a payment can clear every usual application and approval check even when the record behind it has been altered without permission. It asked organisations to verify the underlying business records and the authority for a payment, not just the payment request.

What are AI 'agentic offensive frameworks'?

CERT-In uses the term for automated tools that run several intrusion tasks at once with little manual effort. It said the activity it observed included automated reconnaissance of exposed systems, rapid preparation and modification of exploits for known vulnerabilities, multiple agents working in parallel on reconnaissance and privilege discovery, scripts that change their steps when blocked or when they hit errors, and the use of valid credentials and legitimate administrative functions to go deeper into networks.

The advisory said such frameworks can shorten the time between scouting a target and later stages of an attack, and that every sector should take them into account when reviewing exposure, access controls and monitoring. The concern about AI-assisted intrusions is not limited to India: on 19 November, the European Commission confirmed an AI-assisted cyberattack on its staff systems.

A run of related warnings

The new advisory builds on earlier ones listed on CERT-In's 2026 advisory page. These include a 9 October warning about threat actors exploiting payment and disbursement APIs for unauthorised fund transfers at banking and financial entities, a 15 October advisory on AI agents being used as attack vectors, and a 4 November advisory on an escalation of AI-driven attacks on financial entities by a threat actor it called "CRAFTY FALCON". A Reuters report on 20 November noted that the Reserve Bank of India has tightened cybersecurity checks after a series of attacks on banks and non-bank lenders.

What organisations are being asked to do

  • Government bodies and service owners: identify which provider systems and accounts can change payment records or approve transactions, check payment instructions against an independent record that cannot be altered from the same provider environment, and require independent verification for high-risk changes such as new beneficiaries.
  • Technology service providers and payout platforms: keep client environments separate, restrict administrator and control-plane access, and monitor changes to payment records together with privileged account activity and approvals.
  • All organisations: review internet-facing applications, identities, privileged accounts and service credentials, and examine logs for unusual reconnaissance, credential use and outbound data transfers.

Why it matters and what happens next

The advisory widens the warning beyond banks to the vendors and platforms that sit behind government and corporate payouts, where a single compromised provider could affect several clients. CERT-In reminded organisations that its directions of 28 April 2022 require applicable cyber incidents to be reported to it within six hours of being noticed. Victims of unauthorised financial transactions are advised to report promptly on the National Cyber Crime Reporting Portal or by calling the 1930 helpline, and to inform law enforcement, regulators and payment-system authorities.

CERT-In has not said whether investigations into the reported incidents are complete, and no losses or culprits have been officially identified so far.

Photo source