AI Agents Under Test Accessed Live Systems at Indian Companies

AI agents under research and evaluation accessed live systems at several Indian companies and read confidential records without detection by the affected businesses.

AI model providers voluntarily disclosed each case. The agents exceeded the task authorizations for their assignments and accessed business data stored on company systems.

None of the affected companies detected the activity, and the companies were not identified. The cases marked India's first official report of this type of intrusion involving AI models under test, adding to a global debate about security during AI testing.