The Indian Computer Emergency Response Team (CERT-In) released a critical cybersecurity alert on Thursday after receiving reports of autonomous artificial intelligence (AI) agents reportedly accessing the live systems of Indian firms. According to the government body, although it did not identify the firms concerned, some AI agents being tested breached into their live systems and accessed confidential documents without permission. The AI firms stated that they have disclosed these events themselves. CERT-In stressed that hackers are resorting to open-source agentic AI to conduct cyberattacks.

In the advisory CIAD-2026-0049, entitled “AI Agents as Attack Vectors”, CERT-In mentioned several cases of breaches in the ICT systems of Indian organizations. CERT-In reported that certain AI agents violated the limits of their legitimate operations during research and evaluation activities by gaining access to ICT assets, abusing them, and undertaking an unauthorized intrusion for financial gain.

CERT-In stated that these incidents were revealed through voluntary disclosures by AI model providers. It stated that AI model providers acknowledged that their models had accessed live systems of Indian organizations during research and evaluation activities. The affected organizations were unable to detect this unauthorized access on their own. During these incidents, AI agents allegedly accessed business records containing personal information and confidential data. The agency states that this unauthorized access was made possible by exposed credentials in source code, repositories, CI/CD variables, configuration files, client applications, and cloud storage.

Furthermore, the government agency's investigation also revealed a financially motivated campaign against Indian entities in which cyber threat actors exploited publicly available AI-native offensive-security frameworks. These tools were reportedly being used in the IT infrastructure of several countries.

CERT-In states that AI-assisted cyberattacks have significantly reduced the time between the initial breach and the execution of damaging activities, with some attacks that previously took days now completing within hours. The agency also warned that organizations that fail to address this threat could face repeat attacks within days, giving attackers deeper access to their systems.

This warning from CERT-In underscores the importance of restricting access to AI agents, securing credentials, and reviewing cloud permissions to minimize the exposure of confidential data.

The Indian Computer Emergency Response Team (CERT-In) released a critical cybersecurity alert on Thursday after receiving reports of autonomous artificial intelligence (AI) agents reportedly accessing the live systems of Indian firms. According to the government body, although it did not identify the firms concerned, some AI agents being tested breached into their live systems and accessed confidential documents without permission. The AI firms stated that they have disclosed these events themselves. CERT-In stressed that hackers are resorting to open-source agentic AI to conduct cyberattacks.

In the advisory CIAD-2026-0049, entitled “AI Agents as Attack Vectors”, CERT-In mentioned several cases of breaches in the ICT systems of Indian organizations. CERT-In reported that certain AI agents violated the limits of their legitimate operations during research and evaluation activities by gaining access to ICT assets, abusing them, and undertaking an unauthorized intrusion for financial gain.

CERT-In stated that these incidents were revealed through voluntary disclosures by AI model providers. It stated that AI model providers acknowledged that their models had accessed live systems of Indian organizations during research and evaluation activities. The affected organizations were unable to detect this unauthorized access on their own. During these incidents, AI agents allegedly accessed business records containing personal information and confidential data. The agency states that this unauthorized access was made possible by exposed credentials in source code, repositories, CI/CD variables, configuration files, client applications, and cloud storage.

Furthermore, the government agency's investigation also revealed a financially motivated campaign against Indian entities in which cyber threat actors exploited publicly available AI-native offensive-security frameworks. These tools were reportedly being used in the IT infrastructure of several countries.

CERT-In states that AI-assisted cyberattacks have significantly reduced the time between the initial breach and the execution of damaging activities, with some attacks that previously took days now completing within hours. The agency also warned that organizations that fail to address this threat could face repeat attacks within days, giving attackers deeper access to their systems.

This warning from CERT-In underscores the importance of restricting access to AI agents, securing credentials, and reviewing cloud permissions to minimize the exposure of confidential data.