According to government sources, companies are required to strengthen their cybersecurity measures and intensify monitoring of unusual system activity following the Cert-In warning regarding autonomous artificial intelligence agents. It was noted that critical information of significant organizations was not compromised, and this incident should serve as an early warning.
India's cybersecurity agency, Cert-In, issued a 'critical' warning that autonomous AI agents gained unauthorized access to the work systems and confidential data of Indian organizations. Furthermore, attackers are actively using open-source AI agent tools to conduct rapid cyberattacks.
Government sources informed PTI that upcoming AI regulation, for which a consultative document may soon be released, may also address issues related to autonomous agents. Sources emphasized that the incident did not affect large organizations and did not result in significant financial losses or compromise of vital information, but it highlighted existing vulnerabilities.
Consequently, companies must immediately improve their cybersecurity hygiene, invest in more advanced protection tools, and enhance real-time monitoring levels. Larger enterprises should utilize Security Operations Centers to track suspicious activity, including multiple login attempts and access from unusual IP addresses. Additionally, organizations must implement machine learning tools for threat detection and automated response.
Sources also warned that as AI systems become increasingly capable of performing tasks independently, the possibility that agents might repeat actions without stopping cannot be ruled out. It was stressed that organizations need to improve their ability to analyze system logs, identify atypical behavior, and respond promptly to emerging threats.
Under the critical warning (CIAD-2026-0049) titled 'AI Agents as Attack Vectors', the Indian Computer Emergency Response Team (Cert-In) reported that AI model developers voluntarily disclosed that autonomous AI models and agents, currently in research and evaluation stages, exceeded their authorized scope of tasks and accessed the work systems of Indian organizations.
The agency also noted that attackers used open-source AI agent tools to carry out rapid cyberattacks against Indian entities. It is important to note that none of the affected organizations discovered this access themselves; the incident only became known after the AI model developer reported it following an internal audit.
During these operations, AI agents read business records containing personal data and confidential material by utilizing credentials and connection settings that remained unsecured in application configurations, public code repositories, and overly permissive cloud service roles. In a separate trend identified by Cert-In's own threat hunting, it was observed that cybercriminals are using publicly available AI-based offensive security frameworks for reconnaissance, exploiting security vulnerabilities, and conducting financially motivated campaigns against Indian entities.
These tools continue to be repackaged and redeployed, with many deployments observed on servers in various countries. As part of this campaign, agent tools have reduced the time interval between initial access and final impact from several days to several hours. The agency pointed out that victims who only partially addressed weaknesses were subjected to repeated attacks over several days, but on a larger scale.
Cert-In indicated that the root causes of all recent incidents involving AI agents are traditional security oversights. Jadip Singh, General Manager of Kaspersky in India, stated that this is a 'timely reminder that AI agents are no longer just productivity tools.' He added that with broad access, they can reach work systems and confidential data in ways that organizations never anticipated, often moving through existing gaps such as misconfigured cloud services, exposed credentials, and excessive permissions.
Sunil Sharma, Managing Director and VP of Sales at Sophos (India and SAARC), noted that AI agents are creating new aspects of cybersecurity as they gain the ability to autonomously access data and perform tasks. He advised organizations to extend existing security practices to these systems by establishing separate identification for each agent, implementing the principle of least privilege, securing API keys and tokens, and ensuring adequate logging and continuous monitoring of unusual behavior. Research and testing environments should also be isolated from production ones.
Soumen Mandal from Counterpoint Research speculated that the risk of AI agents accessing confidential information beyond their authorized scope without organizational knowledge could affect ordinary users, as AI agents increasingly process personal and sensitive information. He called on AI developers, governments, application providers, independent researchers, and industry organizations to collaborate in creating robust security standards and accountability frameworks that strengthen trust without slowing down innovation.