CERT-In issues critical alert on AI-powered cyberattacks targeting banks, fintech firms

India’s cyber agency CERT-In has issued a critical-severity advisory on a growing campaign of AI-assisted attacks on banks, lenders and payment firms. It warns that the threat is active and still escalating. The advisory, CIAD-2026-0058, was issued on November 4. It updates an October 9 advisory (CIAD-2026-0046) on attackers exploiting payment and disbursement APIs to make unauthorised fund transfers. CERT-In said continued analysis of reported incidents showed a significant escalation since then.Who is being targetedThe affected entities include non-banking financial companies (NBFCs), digital lending platforms, fintech firms, payment aggregators and payment service providers. Cryptocurrency exchanges and virtual digital asset service providers are also on the list.CERT-In’s financial-sector response team, CSIRT-Fin, has named the attacker “CRAFTY FALCON.” It describes the actor as AI-assisted, one that surveys targets with machine intelligence, works at machine speed and strikes by calling the victim’s own payment and disbursement APIs.Repeat attacksCSIRT-Fin continues to receive reports of successful attacks. These include entities that had received earlier restricted advisories and had already carried out remediation. Some firms were deliberately attacked again after incomplete clean-ups, and attacks were timed for holidays, weekends and other low-monitoring windows to delay detection.AI tools behind the attacksCERT-In said the attackers are using several publicly available AI offensive-security tools, including the ARTEX family of LLM-based tools, the PentAGI autonomous multi-agent penetration testing system, the Cairn attack search engine and the CyberStrikeAI framework. The agency said these tools run complete attack chains on their own, generate and refine large numbers of tailored scripts in quick succession, and adapt payloads based on how the target responds. The attackers’ infrastructure changes IP addresses within hours or days.How the attacks workThe advisory lists these patterns:* Cloud abuse: Attackers misuse cloud management and serverless services, such as AWS Systems Manager and Lambda, to run commands and deploy malicious files, blending in with legitimate activity.* Weak entry points: Internet-exposed test and staging environments are used as a way in, then attackers move to production using shared credentials, tokens or network paths.* Tampered apps: Attackers reverse-engineer application code and send unauthorised API calls using modified app packages, defeating controls that assume only the genuine app will call the APIs.* Allow-list bypass: Traffic is routed through shared third-party fetcher services, such as content-acceleration and app-store review fetchers, to get around IP allow-lists.* Test transactions: Attackers send micro-value payments, such as Rs 1, to check that unauthorised fund flows work before moving large sums.* Settlement APIs: Merchant settlement APIs at payment aggregators are abused to move money.* Fast exploitation of new flaws: Newly published vulnerabilities, including file-upload and deserialisation flaws, are exploited as soon as they become public, leading to web shells in PHP and Node.js applications.* Crypto theft: Wallet keys are stolen from crypto exchanges and funds siphoned out.* AI-generated scripts: Large numbers of near-identical custom scripts appear in quick succession, a pattern CERT-In says is consistent with AI-generated code.What firms are told to doCERT-In asked financial entities to treat the campaign as an immediate and continuing threat. Its recommendations include:* 24x7 monitoring with extra staffing and lower alert thresholds on holidays and weekends.* Assuming attackers persist after any incident, with full forensic clean-up and regular threat hunting.* Strict separation of test and staging environments from production.* Server-side validation of every transaction, API request signing and app attestation, rather than trusting client apps.* Replacing IP allow-listing of shared fetcher pools with identity-based controls such as mutual TLS or signed requests.* Flagging micro-value test payments to new beneficiaries followed by large transfers, and holding or reviewing those flows.* Maker-checker approval and reconciliation for merchant settlement instructions.* Hardware security modules or cold storage and multi-signature approvals for crypto wallet keys.* Banning unauthorised AI offensive tools on corporate systems and watching for LLM API traffic from servers.* Rapid patching of internet-facing applications, honeypots and canary credentials as early-warning measures, and zero-trust architecture.You use AI every day. Now get your AI Quotient. Take the AIQ test.