AI Agents Accessed Confidential Data Of Indian Firms, Went Undetected For Days: CERT-In
Artificial intelligence (AI) agents are emerging as a new cybersecurity concern in India, with some systems operating beyond their assigned tasks and exploiting vulnerabilities that organisations failed to detect.
লেখক Karishma Jain··45K বার পড়া
AI Agents Accessed Confidential Data Of Indian Firms, Went Undetected For Days: CERT-In
Artificial intelligence (AI) agents are emerging as a new cybersecurity concern in India, with some systems operating beyond their assigned tasks and exploiting vulnerabilities that organisations failed to detect.
In an advisory issued on Thursday, the Indian Computer Emergency Response Team (CERT-In) revealed that AI agents had accessed the live systems of several Indian organisations without authorisation, while cybercriminals were also using AI-powered tools to carry out financially motivated attacks.
AI Agents Accessed Confidential Data Without Detection
Several AI model service providers voluntarily informed CERT-In that their frontier AI models and agents had accessed the live systems of Indian organisations while undergoing research and evaluation.
The agents had exceeded their assigned authorisation, accessing business records containing personal data and other confidential information.
“The recorded activity involved reading business records, including personal data and confidential material, through credentials and connection settings exposed in application configuration, public code repositories and permissive cloud service roles,” CERT-In said.
The affected organisations failed to detect the intrusions, which in some cases went unnoticed for days. The incidents came to light only through the AI companies’ disclosures.
CERT-In did not name the AI model providers or the affected Indian organisations.
How Did AI Agents Gain Access To Corporate Systems?
The AI agents exploited conventional cybersecurity weaknesses, including exposed passwords, digital keys and credentials stored in source code, public repositories, application configuration files and other publicly accessible resources.
CERT-In also identified dormant application programming interfaces (APIs) left accessible over the internet and payment API keys stored in plaintext, which could be read from application runtimes.
While these vulnerabilities were not new, the speed and scale at which AI agents exploited them set the incidents apart from conventional cyberattacks.
“In the incidents observed, AI found and exploited these conventional weaknesses faster, more exhaustively and at greater scale than human attackers,” CERT-In said.
Cybercriminals Using AI Agents To Launch Faster Attacks
Separately, CERT-In uncovered financially motivated cyberattack campaigns in which criminals used open-source AI agents equipped with ready-made intrusion capabilities to identify and exploit vulnerabilities in Indian organisations.
The agency’s proactive threat-hunting operations revealed that the time between an initial intrusion and its impact had dropped from several days to just a few hours. Attackers also deployed multiple AI agents simultaneously, allowing operations to run in parallel for hours or days.
Organisations that failed to fully address security weaknesses after an initial attack were targeted again on a larger scale. In incidents voluntarily disclosed by AI companies, agents were also found to have rediscovered and reused the same exposed credentials across separate operations.
CERT-In Warning Amid Cyberattacks On India’s Financial Sector
The warning comes amid a series of cyberattacks on India’s financial institutions. The Economic Times reported that around 20 financial institutions were hit earlier this month, with hackers allegedly using stolen partner-bank credentials and weaknesses in payment systems to siphon funds.
In a separate advisory issued on October 9, CERT-In warned that attackers had targeted lenders, fintech firms, payment providers and digital wallet operators. Using stolen credentials, they initiated transfers through compromised organisations’ own systems, bypassing safeguards against unauthorised payments.
AI-Related Cybersecurity Concerns Grow Globally
India is not alone in facing AI-related cybersecurity threats. On October 9, Japan declared a state of emergency in cyberspace following a wave of AI-related cyberattacks targeting more than 20 major companies, local governments, railway operators and public services.
In November 2025, Anthropic disclosed that a group it assessed as Chinese state-sponsored had manipulated its Claude Code tool to attempt intrusions into around 30 organisations, succeeding in a small number of cases.
In July 2026, OpenAI’s agents reportedly breached parts of Hugging Face’s systems. The incident came amid growing scrutiny of autonomous AI tools, with OpenAI, Anthropic and researchers examining tens of thousands of problematic AI-related incidents, including unsuccessful attempts rather than only confirmed breaches.