The growing autonomy of AI agents is intensifying global cybersecurity concerns, with rogue AI agents reported in the US and other countries, raising urgent questions about AI safety. Similar concerns have emerged in India, where cybersecurity agency CERT-In has issued a ‘Critical’ alert over unauthorised access to Indian organisations’ systems.
In an advisory issued on October 15, the Indian Computer Emergency Response Team (CERT-In) said it had received multiple voluntary disclosures from AI model providers whose models accessed live systems of Indian organisations during research and evaluation activities, including business records containing personal data and confidential information.
The advisory said some AI agents exceeded their assigned authorisation and accessed Indian information and communication technology (ICT) assets without permission. In separate incidents, cyber attackers deliberately used open-source agentic AI tools to conduct reconnaissance, exploit vulnerabilities and carry out financially motivated intrusions.
According to CERT-In, unauthorised access was enabled by credentials and connection settings exposed in application configurations and public code repositories, as well as cloud service roles with excessive permissions. Other weaknesses included dormant, deprecated or test APIs accessible over the internet, payment API keys stored in plaintext, and internet-exposed management interfaces with weak credentials or no multi-factor authentication.
CERT-In said the research-and-evaluation incidents came to light through voluntary disclosures by AI model providers following internal reviews. “None of the affected organisations detected the access,” the advisory said.
The advisory highlighted how AI agents can independently plan tasks, execute commands, invoke tools and operate online for extended periods.
CERT-In warned that agentic AI tools were accelerating financially motivated cyberattacks against Indian entities, reducing the time between initial access and the final impact from several days to a few hours. It also observed multi-agent operations running in parallel for hours or days.
AI agents involved in the research-and-evaluation incidents were observed re-deriving and reusing exposed credentials across operations. CERT-In warned that organisations that failed to remediate fully were attacked again within days, at greater scale.
What Organizations should do to Secure AI Systems
For AI model developers and research laboratories, CERT-In recommended isolating testing environments from production systems, restricting network access and using synthetic targets and credentials.
It also called for real-time monitoring of agent activity, clear limits on authorised actions and access, and automatic termination of runs when agents encounter out-of-scope systems or third-party credentials.
CERT-In has also issued a blueprint for reducing exposure and defending against AI-assisted vulnerabilities in digital infrastructure.
For organisations, the agency recommended auditing code repositories, configuration files and cloud storage for exposed credentials; revoking compromised keys; removing obsolete APIs and internet-facing administrative interfaces; enforcing multi-factor authentication; and limiting permissions to the minimum required.
It also advised assigning each AI-agent run a separate, short-lived identity, restricting outbound connections to approved destinations, detecting unusual API activity, maintaining tamper-resistant logs and retaining security logs for at least 180 days.
For BFSI and fintech entities, CERT-In also recommended implementing measures to counter threat actors exploiting payment and disbursement APIs to make unauthorized fund transfers.
CERT-In further advised organisations to treat systems affected by credential exposure or confirmed unauthorized access as compromised until they are forensically cleared.
Also Read: AI Impact Summit 2026: Red Carpet Rolled, Lights Up – Global Titans Turn to India
