92 pc of Indian organisations face agentic AI security incidents: Study
The study, titled Deployed but Ungoverned: The State of Agentic Security in India, found significant gaps in security governance, visibility and control of AI agents despite their increasing adoption by enterprises.
Based on a survey of 500 security leaders in India, the report found that only 19 per cent of organisations had a designated individual responsible for agentic AI security governance, while just 18 per cent had a unified interface to monitor all their AI agents.
Only 30 per cent had automated security checks before deploying AI agents, while 47 per cent could produce a complete security evidence package on demand, indicating gaps in audit readiness and accountability.
Although 51 per cent of organisations had automated systems to detect abnormal AI agent behaviour in real time, 67 per cent were unable to respond automatically to such incidents, highlighting the gap between threat detection and response.
Kulmeet Bawa, Group Vice-President and Managing Director for India and SAARC at ServiceNow, said AI agents had become part of enterprise operations, but security measures had not kept pace with their adoption.
He stressed the need for organisations to maintain a centralised inventory of AI agents, assign clear accountability and enforce security controls automatically.
The report further found that only 46 per cent of organisations involved security teams before beginning the development of AI agents. While 83 per cent were deploying AI agents in, or interacting with, operational technology or physical environments, only 29 per cent had separate governance frameworks for such deployments.
The growing problem of unregulated AI deployment, known as ‘Shadow AI’, was also highlighted, with 37 per cent of organisations reporting that they had discovered AI agents operating in production environments without being registered or known to their security teams during the past 12 months.
The study noted that the Indian Computer Emergency Response Team’s (CERT-In) October 2026 advisory on AI agents had brought their visibility and control into focus.
It also referred to CERT-In’s 2022 directions requiring organisations to report specified cyber security incidents within six hours of becoming aware of them.
Organisations using unified platforms for agentic AI security, accounting for 43 per cent of those surveyed, reported better security capabilities.
About 62 per cent of these organisations had automated real-time alerts for abnormal AI agent behaviour, compared with 43 per cent of those without such platforms.
Similarly, 45 per cent could automatically address policy violations, against 24 per cent of other organisations, while 51 per cent continuously captured security evidence, compared with 43 per cent.
The study was conducted by ServiceNow in association with Phronesis Partners and covered more than 3,700 senior security and technology decision-makers across 20 countries, including 500 security leaders in India.
The Asia-Pacific region accounted for 1,201 respondents.
