India AI model breach: CERT-In flags live system access
India's cyber agency says AI agents under test entered live company systems, accessed business records, and were only reported by the model providers themselves.
CERT-In said AI models in testing entered live systems of Indian organisations and read confidential records without detection.
Model providers voluntarily disclosed that their agents “exceeded the authorisation of their assigned tasks” during test runs.
Affected firms did not spot the intrusions; the AI companies flagged every case. Providers and victims were not named.
Separately, CERT-In found attackers using freely available AI tools to probe and break into Indian systems for financial gain.
Such attacks that once took days can now unfold in a few hours, with incomplete clean-ups followed by deeper re-attacks within days.