AI agents accessed Indian systems without authorisation, accessed confidential data: CERT-In

Artificial intelligence agents accessed live systems of Indian organisations without authorisation and read business records containing personal data and confidential information, India's cybersecurity agency CERT-In said in an advisory.

The agency said on Thursday that AI models operating during research and evaluation had exploited exposed credentials and permissive cloud settings to access the systems.

The affected organisations failed to detect the unauthorised access, and the incidents came to light through voluntary disclosures by AI model providers after internal reviews.

CERT-In also reported financially motivated cyberattacks in which threat actors used publicly available AI-based offensive security tools for reconnaissance, exploiting vulnerabilities and intrusions.

The use of such tools had reduced the time between initial access and the final impact of attacks from days to a few hours, the agency said in the advisory dated October 15.

"In the incidents observed, AI found and exploited these conventional weaknesses faster, more exhaustively and at greater scale than human attackers," CERT-In said.

The agency found that attackers deployed multiple AI agents in parallel for hours or days to achieve their objectives. Organisations that incompletely addressed vulnerabilities were targeted again within days on a larger scale.

The vulnerabilities included credentials and keys exposed in source code, public repositories and configuration files; dormant or test application programming interfaces (APIs) accessible over the internet; payment API keys stored in plaintext; excessive cloud permissions; and internet-facing administrative interfaces with weak credentials or no multi-factor authentication.

CERT-In advised AI model developers and research laboratories to isolate testing environments from production systems, restrict network access to explicitly permitted destinations and use synthetic credentials and data instead of real business information.

It also called for real-time monitoring of AI agent activity and “if an agent accesses a real system, notify the affected organization and CERT-In promptly.”

The agency also recommended automated monitoring for unusual API activity and retaining logs for at least 180 days. It urged organisations to treat environments as compromised until forensic checks cleared them after credential exposure or confirmed unauthorised access.

Banking, financial services and insurance companies, as well as fintech firms, “should also implement the payment and disbursement API measures” set out in an earlier CERT-In advisory, it said.