The AI that companies considered helpful has now become a major headache for them. AI agents have secretly infiltrated the systems of Indian companies. CERT-In has issued a warning.
AI Agent Cyber Attack
Artificial intelligence (AI) is being used everywhere these days. Many large companies are also leveraging it. Autonomous AI agents are being deployed in everything from coding to data management. But the very technology that makes our work easier now seems to be making it difficult. India's cybersecurity agency, the Indian Computer Emergency Response Team (CERT-In), recently issued a warning.
This has given the corporate world sleepless nights. It has been revealed that during testing and research, some AI agents overstepped their bounds and entered the live systems of Indian companies. They accessed confidential files and private records without permission. Most alarmingly, the companies whose systems were compromised were unaware of this. Later, the AI model developers themselves came forward and revealed this.
AI models have crossed their limits
CERT-In has revealed the full extent of this matter in its advisory CIAD-2026-0049. According to the agency, some AI agents were deployed solely for research and evaluation purposes, but they exceeded their mandate. These agents infiltrated the IT and communication networks (ICT infrastructure) of Indian institutions and extracted business records containing personal information and highly sensitive data. The question arises: how did these agents gain access?
In fact, the companies' own technical negligence was also responsible for this. Passwords and login credentials were exposed in various places across source code, repositories, CI/CD variables, configuration files, client apps, and cloud storage. AI agents detected these weak links and directly accessed live systems. Although the government has not revealed the names of the affected companies, this incident has shaken their cybersecurity capabilities.
Open-source tools become hackers' new weapon
The danger isn't limited to AI models being tested going astray. Cybercriminals are now exploiting this technology for profit. A CERT-In investigation has revealed that a network of hackers is targeting Indian institutions for financial gain. To this end, cybercriminals are blatantly abusing open-source agentic AI tools and offensive-security frameworks readily available online.
Similar attacks are occurring on IT systems in India as well as in many other countries. Most worryingly, the introduction of AI has exponentially increased the speed of cyberattacks. Previously, hackers would take days to breach a company's network and cause damage, but with AI, this task is now accomplished in just a few hours. This means that by the time the security team reacts, the game is over.
Take these steps immediately to protect yourself
The government agency has warned that companies that take this threat lightly and fail to address their flaws in a timely manner could face repeated attacks within days. After an attack, hackers gain access to even deeper parts of the network, potentially causing even greater damage. To mitigate this threat, CERT-In has advised companies to immediately strengthen their security.
The first step is to avoid giving any AI agent open access to the system and set a time limit for their work. Also, keep all your passwords and credentials safe. Don't leave them open in Coding Files, and regularly review cloud storage permissions.
