Q-Cyber Magazine · Advisory

BFSI · 23 November 2026 · 7 min read · Updated

CERT-In names "CRAFTY FALCON": AI-assisted attacks on BFSI payment and disbursement APIs

CERT-In's critical advisory CIAD-2026-0058 of 4 November names CRAFTY FALCON, an AI-assisted threat actor that abuses BFSI entities' own payment and disbursement APIs, and sets out controls.

High-rise towers on a city skyline under a hazy evening sky

New Delhi, 23 November 2026. The Indian Computer Emergency Response Team (CERT-In) has issued a critical-severity advisory warning banking, financial services and insurance (BFSI) entities of an escalating, AI-assisted campaign that moves money by calling the targets' own payment and disbursement APIs. Advisory CIAD-2026-0058, issued on 4 November, says the threat is "active and ongoing" and names the actor behind it "CRAFTY FALCON".

What the advisory says

According to CIAD-2026-0058, continued analysis of reported incidents "has revealed a significant escalation" of the campaign first described in October. CERT-In says it has identified the use of several publicly available AI-driven offensive security tools against affected entities, naming the ARTEX family, PentAGI, Cairn and CyberStrikeAI. It says these tools run complete attack chains autonomously, generate and refine large numbers of tailored attack scripts and adapt their payloads to the target's responses, while the supporting infrastructure rotates IP addresses within hours to days.

The advisory describes the attack pattern seen in the reported incidents:

  • Cloud management channels. Cloud compute, serverless and management services were targeted, with services such as AWS Systems Manager and Lambda abused to run commands and deploy malicious files.
  • Non-production as the way in. Internet-exposed UAT, staging and testing environments were used as entry points, followed by movement to production through shared credentials, tokens or network paths.
  • Modified client applications. Attackers reverse-engineered application code and made unauthorised API calls using altered application packages, "defeating controls that assumed only the legitimate client application would call the APIs".
  • Allow-listing bypassed. IP allow-lists were bypassed by sending traffic through whitelisted third-party fetcher IP pools, such as content-acceleration and app-store review services.
  • Test payments first. Micro-value transactions, such as Rs. 1, were used to validate unauthorised fund flows before large-value transfers.
  • Settlement and wallets. Payment aggregators' merchant settlement APIs were abused to move funds, and at crypto exchanges wallet keys were stolen and funds siphoned.
  • Fast exploitation. Newly published vulnerabilities were exploited as soon as they became public, including file-upload and deserialisation chains leading to web shells in PHP and Node.js applications.

CERT-In adds that attacks have been timed for holidays, weekends and other low-monitoring windows, and that large numbers of rapidly refined, functionally identical scripts are "consistent with AI-generated code".

What came before

The October advisory, CIAD-2026-0046, described campaigns against NBFCs, digital lending platforms, fintech entities, payment service providers and wallet operators. It said attackers stole payment aggregator, gateway and partner-bank API credentials and called disbursal and fund-transfer APIs "directly from within the entities' own whitelisted infrastructure", bypassing application-level checks such as KYC, sanction, beneficiary verification and per-transaction limits. It reported dummy reference identifiers, transfers to mule accounts across multiple banks over IMPS and NEFT, and log deletion and tampering with endpoint detection tools. It already noted the "possible use of AI-assisted offensive tooling".

On 3 November, the National Cybercrime Threat Analysis Unit of the Indian Cyber Crime Coordination Centre (I4C), in coordination with CERT-In, issued a separate advisory on AI-assisted attacks on NBFCs and fintech platforms. It said that "multi-crore unauthorised transactions through fund disbursement systems and APIs are being found in individual cases", The New Indian Express reported. According to that report and Hindustan Times, the I4C advisory asks institutions to report such incidents to CERT-In within six hours of detection, and banks and NBFCs to report them to the Reserve Bank of India through its DAKSH portal within the same period.

What is stated, and what is not

Stated by CERT-In: the actor designation; the sectors from which incidents were reported; that successful attacks continue to be reported; the tools identified; the attack pattern above; and the recommended actions.

Not stated in the public advisories: which entities were affected, how many incidents there have been, the amounts involved, or who is behind CRAFTY FALCON. CERT-In does not attribute the campaign to a country or group. The full indicators of compromise, infrastructure fingerprints and detection rules are in the restricted version only, and CERT-In warns entities not to limit their response to blocking indicators, because "indicators may change while adversary techniques often remain consistent".

What CERT-In recommends

The 27 actions in CIAD-2026-0058 fall into five groups:

  1. Monitoring and response. Maintain heightened 24x7 monitoring, with more staff and lower alert thresholds during holidays and weekends. After any incident, assume persistence, carry out full eradication with forensic verification and keep hunting. Tune SIEM correlation for machine-speed, multi-technique attacks and adopt automated containment.
  2. Cloud and environments. Monitor cloud audit logs in real time, restrict cloud management and serverless execution to approved principals, enforce least-privilege roles and no default internet egress for serverless functions. Isolate UAT, staging and testing from production, with no shared credentials, tokens, keys, databases or network paths.
  3. Applications and APIs. Treat client applications as untrusted: validate every transaction server-side, sign API requests with a nonce and timestamp, and use app attestation and certificate pinning. Replace IP allow-listing of shared fetcher pools with identity-based controls such as mutual TLS, signed requests or scoped API keys. Remove hard-coded secrets, scan repositories and pipelines, and use short-lived, least-privileged tokens.
  4. Payments. Flag micro-value payments to new beneficiaries followed by high-value transfers, and hold or review such flows. Apply velocity rules, new-beneficiary cooling periods and amount-band anomaly detection, and enforce maker-checker approval and reconciliation for merchant settlement instructions. Keep crypto wallet keys in hardware security modules or cold custody, with multi-signature approval for withdrawals.
  5. AI tools and deception. Prohibit unauthorised attack tools, including AI-driven offensive frameworks, on corporate systems; govern LLM API keys and usage; and require written approval, scope and a time window for all penetration testing, because "the tools themselves do not validate authorization". Deploy honeypots, decoy panels and canary credentials, and adopt zero-trust principles on the assumption that the actor is already inside.

Suspicious activity should be reported to CERT-In and CSIRT-Fin at [email protected] with relevant logs, and fraudulent transactions to I4C through its Cyber Fraud Mitigation Centre.

What it means

  1. The payment API is the target, not the app. In both advisories, controls that lived only in the client application or in earlier workflow stages were bypassed by calling the API directly. Checks such as sanction, beneficiary verification and limits need to be enforced where money moves. The articles on RBI's 2026 digital payment security controls and third-party risk in Indian BFSI cover the regulatory side; Knowlcom's cybersecurity audit work reviews the architecture and configuration behind these flows.
  2. Network location is not identity. An allow-listed IP range, a shared fetcher pool or a token in a staging system all failed as proof of who was calling. CERT-In's answer is identity for every user and service, mutual TLS or signed requests, and no shared credentials between environments. See zero-trust architecture and identity and least privilege.
  3. Keys and signatures carry the weight. Request signing, mutual TLS and HSM or cold-custody storage for wallet keys all depend on how keys are generated, stored and rotated. Entities building these controls now can record the algorithms in a cryptographic inventory at the same time, which India's quantum-safe roadmap asks critical sectors to start; see data security, key management and crypto-asset custody.
  4. Defend at machine speed, report at human speed. The advisories describe attacks timed for weekends and scripts refined faster than people can read them. Automated containment, transaction holds on suspicious patterns and rehearsed reporting are all needed; the CERT-In six-hour reporting rule still applies. CERT-In's rules on authorised testing and governed LLM usage also apply to defenders' own AI deployments.

CERT-In describes the campaign as continuing. This article will be updated if CERT-In or CSIRT-Fin publish further advisories.

Sources

  1. CERT-In, Advisory CIAD-2026-0058: Escalation of AI-Driven Attacks on BFSI Entities by Threat Actor "CRAFTY FALCON" (4 November 2026)
  2. CERT-In, Advisory CIAD-2026-0046: Threat Actors Exploiting Payment and Disbursement APIs for Unauthorized Fund Transfers in BFSI Entities (9 October 2026)
  3. The New Indian Express (Mukesh Ranjan), MHA's I4C warns banks, fintech firms of AI-powered cyberattacks and machine-speed fraud (4 November 2026)
  4. Hindustan Times (Sejal Sharma), Cert-In, I4C warn financial firms of AI-driven cyberattacks, unauthorised fund transfers (5 November 2026)
  5. MediaNama (Prabhanu Kumar Das), CERT-In flags an AI threat actor targeting BFSI payment APIs for unauthorized fund transfers (5 November 2026)