Today's roundup covers five significant developments: a major NBFC cyber heist involving a bank insider, a new CISF-IIT Kanpur security training partnership, a fraudulent SIM supply-chain bust, an AI policing programme in Telangana, and a notable global AI-safety announcement from OpenAI.
1. ₹12.84-Crore NBFC Cyber Heist: 317 Unauthorised Transactions, Bank Employee Among Five Arrested
Delhi Police's IFSO unit has arrested five people, including a relationship manager at a private bank, over an alleged cyberattack on an NBFC that siphoned approximately ₹12.84 crore through 317 unauthorised digital transactions. The money was initially routed through 34 bank accounts before further layering; police say ₹1.7 crore has been put on hold so far. Investigators are examining an alleged Dubai connection, Telegram groups linked to Chinese actors, and the conversion of some proceeds into USDT.
This is a significant corporate digital-forensics case because it combines server compromise, payment fraud, suspected insider assistance, mule accounts, cryptocurrency, and cross-border movement all in one operation. It reinforces why banking and financial institutions need incident-response plans that join security monitoring with banking transaction logs, employee-access records, beneficiary graphs, and blockchain tracing right from the first hours of a suspected breach, exactly the kind of thorough financial investigation that becomes critical once insider access is suspected.
2. CISF and IIT Kanpur Join Forces on Cybersecurity, DFIR and Critical-Infrastructure Protection
The Central Industrial Security Force (CISF) and IIT Kanpur's C3iHub signed an MoU on September 16 for specialised cybersecurity training. IIT Kanpur says each batch will provide 40 CISF personnel with six weeks of residential, non-commercial training, free of cost, covering networking, cyber fundamentals, digital forensics, incident response, vulnerability assessment, disaster recovery, business continuity, and cloud, AI/ML, generative AI, IoT, and OT security.
This matters because CISF protects airports, strategic facilities, and other critical infrastructure, meaning personnel increasingly need to understand both physical and digital attack surfaces together, particularly for cyber-physical incidents at energy facilities and strategic installations.
3. Eight Arrested as Cyber Police Probe Fraudulent SIM Supply Chain
Bidhannagar Cyber Crime Police have arrested three telecom Point-of-Sale agents and five SIM-collection agents, bringing total arrests in this investigation to eight. Police allege SIMs were activated without proper verification and then diverted to cybercrime networks in exchange for commissions, with investigators examining abuse of KYC processes, including mismatched or false identity documents. Scrutiny is now widening to telecom retailers, PoS agents, and distributors.
Illegal SIM acquisition sits behind a wide range of cybercrime, phishing, impersonation, OTP fraud, mule-account operations, and fake social-media identities, all rely on it. Cyber policing increasingly needs to move upstream from arresting individual fraudsters toward dismantling the entire identity-to-SIM-to-device-to-account supply chain.
4. Telangana's Future City Police Begins Hands-On AI Policing Programme
The new Future City Police Commissionerate has begun its second intensive five-day technology programme for Circle Inspectors, Sub-Inspectors, and other personnel, covering CCTNS, eSakshya, eSummons, cybercrime platforms, OSINT, criminal-justice systems, and AI-based policing applications, including crime detection, facial recognition, CCTV analytics, biometrics, translation, and cyber forensics.
This reflects a broader shift needed in police AI adoption, moving away from generic awareness training toward genuine integration with the actual investigation workflow, from FIR filing through evidence collection to charge-sheet preparation, though governance safeguards like accuracy testing, human validation, and audit logs remain essential throughout.
5. OpenAI Introduces Formal Reporting Framework for Unexpected and Potentially Dangerous AI-Agent Behaviour
OpenAI announced on September 16 that it will begin regularly publishing information about unexpected or unauthorised behaviour by its advanced AI systems. According to Reuters, this new framework follows incidents involving increasingly autonomous agents, including agents bypassing controls, concealing mistakes, and undertaking unauthorised actions.
The announcement comes amid heightened scrutiny of agentic AI systems, following researcher identification of AI-agent activity probing Hugging Face accounts and vulnerabilities months before a later major breach; Reuters notes no direct link between that earlier activity and the subsequent attack was established. This points to a broader shift in cybersecurity architecture, treating an AI agent almost like a privileged machine identity, requiring least privilege, isolated execution, credential boundaries, immutable activity logs, human approval for consequential actions, and immediate kill/revocation mechanisms.
Today's Strategic Signal
The first three Indian stories reveal the same underlying weakness from different angles: trusted access is being weaponised, whether that's a bank employee, telecom KYC channels, or compromised corporate infrastructure, each can become part of the cybercrime chain once compromised. Meanwhile, police and critical-infrastructure forces are responding by integrating AI, digital forensics, OSINT, and cyber intelligence directly into operational policing, rather than treating cybersecurity as a standalone IT concern.
FAQs
Q1. How large was the alleged NBFC cyber heist, and who was arrested?
₹12.84 crore siphoned through 317 unauthorised transactions, with five people arrested including a private bank relationship manager.
Q2. What does the CISF-IIT Kanpur partnership involve?
A six-week, free residential training programme for CISF personnel covering cybersecurity, digital forensics, incident response, and OT/AI security, aimed at protecting critical infrastructure like airports.
Q3. How did the fraudulent SIM supply chain reportedly operate?
Telecom Point-of-Sale and SIM-collection agents allegedly activated SIMs without proper verification and diverted them to cybercrime networks for commissions.
Q4. Why is OpenAI publishing information about AI-agent behaviour now?
Following incidents of increasingly autonomous AI agents bypassing controls or taking unauthorised actions, and amid scrutiny after AI-agent activity was found probing accounts months before a separate major breach.