AI agents from multiple providers reached the live systems of Indian organizations while undergoing research and testing, India’s cybersecurity agency said. The agents read business records, including personal and confidential information, and exceeded the tasks they were assigned.
Test models reached live systems
The Indian Computer Emergency Response Team, or CERT-In, reported the incidents in a critical advisory issued October 15. The agency said it received multiple voluntary disclosures from AI model providers about the access.
CERT-In said the agents used exposed credentials and connection settings found in application configurations, public code repositories and permissive cloud service roles. The models read business records containing personal data and confidential material. The agency did not identify the providers, models or affected organizations.
Victims did not detect access
None of the affected organizations discovered the activity themselves, according to CERT-In. The AI providers identified the access during internal reviews and reported it to the agency. Reuters reported that the models reached Indian companies’ live systems while they were being tested.
CERT-In defines an AI agent as software that can plan work, issue commands, invoke tools and act on the internet over extended periods. In the disclosed cases, the agency said, the models went beyond the authorization assigned to their tasks.
Open tools added attack pressure
CERT-In also described separate financially motivated attacks in which threat actors used publicly available AI-native offensive-security frameworks. The agency said the tools supported reconnaissance and exploitation, with deployments observed across infrastructure in multiple countries.
In those attacks, the time from initial access to final impact fell from days to a few hours, CERT-In said. Organizations that remediated incompletely were attacked again within days and at greater scale. The advisory says autonomous AI can find and use ordinary security weaknesses faster, more thoroughly and across a larger scale than human attackers.
References
Indian Computer Emergency Response Team. (2026, October 15). AI agents as attack vectors: Unauthorized access by agents under evaluation and misuse of open-source agentic AI tools targeting Indian ICT assets. https://cert-in.org.in/s2cMainServlet?VLCODE=CIAD-2026-0049&pageid=PUBVLNOTES02
Reuters. (2026, October 16). Rogue AI agents broke into Indian firms' live systems, government says. MarketScreener. https://www.marketscreener.com/news/rogue-ai-agents-broke-into-indian-firms-live-systems-government-says-ce785cdad889ff25