Nearly all Indian organisations are grappling with AI agent security failures, with 92 per cent reporting at least one agentic security incident in the past year, according to new research by ServiceNow. The report, Deployed but Ungoverned: The State of Agentic Security in India, surveyed 500 senior security leaders and found that only 36 per cent of affected organisations managed to contain incidents within 24 hours.
The findings point to a stark mismatch between confidence and control. Half of security leaders describe their agentic security posture as “established,” yet only 19 per cent have a single named individual accountable for agentic AI governance, and just 18 per cent have unified visibility across all their AI agents from one interface.
Detection capabilities are ahead of response capabilities. Around 51 per cent of organisations have automated real-time monitoring for unusual agent behaviour, but 67 per cent cannot automatically act when an agent behaves anomalously, leaving human teams to respond manually against threats that move at machine speed.
A significant blind spot compounds the problem. Some 37 per cent of organisations discovered AI agents running in production over the past year that were never registered with or known to security teams, a phenomenon commonly described as “shadow AI.”
The exposure extends beyond digital infrastructure. Some 83 per cent of organisations run AI agents in or connected to operational technology or physical environments, yet only 29 per cent have a separate governance framework covering those deployments. Only 46 per cent involve security teams before agent development begins.
Organisations running agentic security on a unified platform show meaningfully better outcomes: 62 per cent have automated real-time alerts compared to 43 per cent among others, and 45 per cent can automatically remediate policy violations versus 24 per cent.
The findings come against the backdrop of a CERT-In advisory on AI agents issued in October 2026 and existing rules requiring Indian organisations to report cyber incidents within six hours of detection.
The global study covered more than 3,700 security and technology decision-makers across 20 countries, conducted in partnership with Phronesis Partners.
Published on November 18, 2026




