In its advisory titled ‘AI Agents as Attack Vectors’, the cyber agency said threat actors are actively weaponising open-source agentic AI tools to carry out rapid cyberattacks. The advisory highlighted two distinct categories of incidents targeting the Information and Communication Technology (ICT) infrastructure of Indian entities.
In one incident, AI agents accessed business records containing personal information and confidential data. In another, cybercriminals used AI-powered tools to identify security gaps and launch financially motivated attacks against Indian organisations.
According to CERT-In, the AI agents gained access to business records by exploiting credentials and connection settings exposed through application configurations, public code repositories and cloud service roles with excessive permissions. Notably, none of the affected organisations detected the unauthorised access themselves, PTI reported.
AI Tools Reduce Cyberattack Timelines from Days to Hours
CERT-In warned that AI agents have significantly reduced the time between gaining initial access to a system and causing damage. In the campaign identified by the agency, this window shrank from several days to just a few hours.
The agency also reported that organisations that only partially addressed security weaknesses were attacked again within days, with subsequent attacks carried out on a larger scale.
CERT-In found that basic security failures were responsible for all the recent AI agent-driven incidents it observed. These included passwords, access keys and other credentials exposed in source code, public repositories, configuration files, client applications and publicly accessible web content.
Other vulnerabilities included outdated or test application programming interfaces (APIs) that remained accessible over the internet, payment API keys stored in plain text, cloud systems configured with excessive permissions and management interfaces exposed to the internet.
Unpatched software, weak passwords and the absence of multi-factor authentication further increased the vulnerability of affected systems. CERT-In noted that AI agents could identify and exploit these weaknesses faster and on a larger scale than human attackers.
CERT-In Issues Security Guidelines for Companies
To address these risks, CERT-In has advised organisations to inspect their source code, public repositories and commit histories, configuration files, client applications and cloud storage for exposed credentials.
Companies have been instructed to change compromised or exposed credentials and remove them from code histories. The agency has also recommended shutting down unused, outdated and test APIs, restricting cloud access to only what is necessary, removing administrative interfaces from the public internet and enforcing multi-factor authentication.
Organisations have further been advised to monitor their systems for attacks and retain logs for at least 180 days.
CERT-In said companies should treat their systems as compromised following a credential leak or confirmed intrusion until a forensic investigation establishes that they are clear. Organisations should also prepare for repeat attack attempts within days or weeks.
AI Developers Advised to Isolate Research Environments
The advisory also outlined precautions for AI developers, research laboratories and evaluation agencies. CERT-In recommended isolating research and testing environments to prevent AI agents from accessing sensitive production systems and data.
Outbound network access should be blocked unless specifically required for a task. The agency stressed that these restrictions must be enforced at the network level rather than relying on instructions given to the AI model.
It also advised developers and researchers to use synthetic targets and test credentials instead of real production data during research and evaluation.
The incidents highlighted by CERT-In demonstrate how exposed credentials, excessive permissions and inadequately secured systems can provide entry points for AI-assisted attacks. With attack timelines shrinking from days to hours, the agency’s recommendations place emphasis on stronger access controls, continuous monitoring and security measures that limit the ability of AI agents to exploit vulnerable infrastructure.
About The Author
Comment List