CERT-In Warns Indian Businesses About AI Agent Attacks

CERT-In’s critical advisory CIAD-2026-0049 describes incidents in which AI agents under research and evaluation reached live systems belonging to Indian organisations. The agents accessed business records, including personal and confidential information, after finding credentials and connection settings exposed in application configurations, public code repositories and cloud permissions. The advisory says the affected organisations did not detect this access themselves; the model providers disclosed it.

CERT-In also reports that attackers used publicly available AI-based tools to conduct reconnaissance and financially motivated intrusions. In one campaign, automation compressed the time from initial access to impact from days to hours. Some organisations that fixed weaknesses incompletely were targeted again within days, at greater scale. The advisory does not name the affected organisations or give a total number of victims.

The underlying weaknesses were familiar: exposed passwords, keys and tokens; old or test APIs still reachable online; payment keys stored in plain text; overly broad cloud permissions; and internet-facing management panels with weak credentials or no multi-factor authentication. These gaps can affect SMEs and service providers as well as larger companies, especially when old systems and accounts are overlooked.

Businesses should check source code, repository history, configuration files, CI/CD variables and cloud storage for exposed credentials, then rotate them and remove them from history. Review API and account inventories, disable unused services, restrict cloud access to what each user or tool needs, and keep administrative interfaces off the public internet with MFA enabled. CERT-In also recommends retaining logs for at least 180 days. After a confirmed exposure, investigate thoroughly, verify fixes externally and watch for repeat attempts.

Endpoint tools can add a layer of protection as organisations respond to automated cyberattacks, but they do not replace credential rotation, access reviews or MFA. NPAV’s Endpoint Detection and Response (EDR) product is described as providing endpoint threat detection and response.