CERT-In warns AI agents overstepped tests and accessed confidential records in Indian systems

India’s national cyber emergency agency, the Indian Computer Emergency Response Team (CERT-In), issued an advisory this week titled "AI Agents as Attack Vectors," saying AI agents run by unnamed model providers exceeded the scope of tasks granted to them during testing and read confidential business records from online systems used by Indian organizations. CERT-In said none of the affected organizations detected the activity on their own. The agency said it received voluntary disclosures from multiple AI model providers whose models, while being researched and evaluated on the internet, touched online systems belonging to Indian institutions. The notice did not name the providers, the models, or the affected entities. According to CERT-In, the agents accessed business records containing personal data and confidential material by using credentials and connection settings exposed in application configurations, public code repositories, and overly permissive cloud service roles. In some cases, the same credentials were recovered multiple times and reused across different operations. CERT-In also said it had observed criminals using publicly available AI-driven offensive security frameworks to scan for vulnerabilities, exploit them, and launch financially motivated attacks against Indian entities, cutting the time from initial access to impact from days to hours. The tools, it added, have been repeatedly repackaged and deployed on servers across multiple countries.