92% of Indian firms had an AI agent incident in a year
Only 36% of the Indian firms with an incident stopped it in 24 hours or less, says a ServiceNow report.
On 18 November 2026, ServiceNow published its Deployed but Ungoverned report on 500 security leaders in India, a part of a global study in 20 countries. Of them, 92% had an incident with an AI agent, which is software that works without help. Only 36% of the firms with an incident in the last year stopped it in 24 hours or less.
What the report found
Half of the leaders say their control of agents is good. But only 19% have one person for agent rules. Only 18% can see all their agents from one screen.
The numbers
- 51% have automatic real-time monitoring of agents.
- 67% cannot make a system take action on its own when an agent is unusual.
- 37% found agents in production that the security team did not know.
- 83% use agents that touch physical systems, but only 29% have different rules for them.
What is not known
Firms with one platform for all agents did better: 45% can repair policy breaches automatically, and 24% of the others can. ServiceNow sells such platforms, and the report does not show that the platform is the cause.
Insight: Firms see agent problems but cannot respond
More firms monitor than stop agents: 51% have monitors, but 67% cannot take action automatically. ServiceNow data for the UK in October 2026 shows the same result: 88% had an incident, and 44% stopped it in 24 hours or less. ServiceNow also sells platforms for this, so the data seems to help its sales too. Get proof before you buy.
Sources
Posted