India's national cyber agency has flagged AI agents as a critical threat to the country's digital systems, warning that some have already broken into Indian networks.
These networks are either beyond the limits of their assigned tasks or in the hands of attackers.
In an advisory issued on Thursday, the Indian Computer Emergency Response Team (CERT-In) rated the threat "Critical".
It said AI agents, software that can plan work, issue commands and use tools on the internet over extended periods, have been the attacker in several recent incidents.
In one category, AI models under research and evaluation by their developers reached the live systems of Indian organisations, CERT-In said.
The agents read business records, including personal data and confidential material, using credentials exposed in application configuration, public code repositories and permissive cloud service roles.
None of the affected organisations detected the access. Each case came to light only because the AI model providers disclosed it after an internal review, the advisory said.
In the second category, CERT-In's own threat hunting found a financially motivated campaign against Indian entities, in which attackers misused publicly available, AI-native offensive-security tools.
ALSO READ | Hackers Turn To Chinese AI Agent To Target South Korea's Biggest Banks, Raising Security Concerns
It said these tools are being rebranded and have been deployed across infrastructure in several countries.
The agency said such tooling cut the time between initial access and final impact from days to a few hours. Victims that fixed problems only partly were attacked again within days, at greater scale.
CERT-In said the underlying weaknesses were conventional: exposed credentials and keys, forgotten or test APIs reachable from the internet, plaintext payment-API keys, over-permissive cloud roles, and exposed management interfaces with weak credentials or no multi-factor authentication.
The difference was that AI found and exploited them faster and at greater scale than human attackers.
ALSO READ | Millions Of Records, Dozens Of Companies: Inside Japan's Escalating Cyber Crisis
The advisory asks AI developers to run evaluations with default-deny network access enforced at the network layer, use synthetic targets and credentials, and notify affected organisations and CERT-In if an agent reaches a real system.
Organisations are advised to audit and rotate exposed credentials, remove dormant APIs, apply least-privilege cloud roles, and keep logs for at least 180 days.
Those deploying their own AI agents should give each run a short-lived identity and keep a tamper-resistant activity log.
Banks, financial services and fintech firms have been directed to also follow payment API measures in an earlier advisory.
Essential Business Intelligence, Sharp Market Insights, Practical Personal Finance Advice, Daily Fuel, Gold and Silver Prices and Latest Stories — On NDTV Profit.