CERT-In: AI agents exploited old misconfigurations in corporate systems

India's CERT-In advisory CIAD-2026-0049 says AI agents under evaluation reached live systems of Indian organizations using exposed credentials, dormant APIs and overly permissive cloud roles. Agentic attack tooling cut the time from initial access to impact from days to a few hours.

• Agents read confidential business records and no affected organization detected the access
• Agentic tooling compressed time from initial access to impact from days to hours
• Incompletely remediated victims were attacked again within days at greater scale
• CERT-In recommends keeping 180 days of logs and requiring MFA on admin interfaces