92% of Indian Firms Hit by AI Agent Incident: ServiceNow Survey

A ServiceNow survey of 500 security leaders found that 92 per cent of Indian organisations had an AI agent security incident in the past year, yet only 36 per cent of them contained it within 24 hours. The report, "Deployed but Ungoverned", points to gaps in ownership, visibility and automated response as enterprises deploy agentic AI faster than they can govern it.

92% of Indian Firms Hit by AI Agent Incident: ServiceNow Survey

Key takeaways

  • 92 per cent of Indian organisations reported an AI agent security incident in the past year; only 36 per cent contained it within 24 hours.
  • Just 19 per cent have a single named person accountable for agentic AI security governance.
  • 37 per cent found unregistered AI agents running in production that security teams did not know about.
  • 51 per cent have automated real-time anomaly detection, but 67 per cent cannot automatically act on anomalous agent behaviour.

Nearly every Indian organisation surveyed has already felt the security fallout of deploying AI agents, but few can respond quickly. A survey of 500 security leaders by ServiceNow found that 92 per cent of Indian organisations had an AI agent security incident in the past year, and only 36 per cent of them were able to contain it within 24 hours, the company said on Wednesday.

What the survey found

The findings are published in ServiceNow's report, "Deployed but Ungoverned: The State of Agentic Security in India". Half of the Indian security leaders surveyed described their agentic security as "established". The data in the report, however, suggests the foundations underneath that confidence are uneven.

Key numbers from the report include:

  • 19 per cent have a single named individual accountable for agentic AI security governance.
  • 18 per cent have unified, single-interface visibility into their AI agents.
  • 30 per cent enforce an automated deployment gate that blocks agents until they pass security checks.
  • 47 per cent can produce a complete evidence package on demand.
  • 51 per cent have automated real-time detection for anomalous AI agent behaviour, but 67 per cent cannot automatically act when an agent behaves anomalously.
  • 37 per cent discovered AI agents running in production in the last 12 months that were never registered or known to security teams.

Detection is ahead of response

One of the sharper contrasts in the report is between spotting a problem and doing something about it. While about half of organisations (51 per cent) have automated, real-time detection for unusual agent behaviour, two-thirds (67 per cent) cannot automatically act when such behaviour is flagged. In practical terms, an alert may reach a security team, but a person must then step in to investigate and respond. That gap can matter when software agents operate continuously and at machine speed.

The 24-hour containment figure points the same way. With only about one in three affected organisations able to contain an incident within a day, a large share take longer to bring an incident under control.

Unknown agents and unclear ownership

The report also highlights a visibility problem. As many as 37 per cent of organisations found AI agents running in production that had never been registered or made known to security teams. Security practitioners generally describe such unmanaged tools as "shadow" deployments: systems that run outside formal oversight and are therefore harder to monitor, audit or switch off.

Accountability appears similarly thin. Only 19 per cent have one named individual responsible for agentic AI security governance, and just 18 per cent can see their agents through a single interface. Without a clear owner and a central view, it can be difficult to answer basic questions such as which agents exist, what they can access and who is responsible if something goes wrong.

What ServiceNow said

Kulmeet Bawa, GVP and Managing Director for India and SAARC at ServiceNow, said: "India's AI ambitions must be matched by the ability to understand, monitor and govern the agents enterprises deploy. Knowing what an agent can reach is not the same as seeing what it is doing. Automated controls and continuous oversight will be essential to keeping AI innovation underpinned by trust and cyber resilience."

He pointed out that AI agents are now part of how Indian enterprises operate, but security has not kept pace. According to Bawa, every agent should be registered, owned and visible in one place, with controls enforced automatically and responses happening at machine speed, so that security teams can detect problems and contain them in time.

Why it matters for investors

For retail investors, the report is a reminder that the adoption of AI inside companies is not only a productivity story. As businesses hand more tasks to autonomous software, operational and cyber risks can grow alongside. Companies that listed firms rely on for customer data, payments or core operations may face disruption, regulatory scrutiny or reputational damage if AI-related incidents are not contained quickly. The survey does not name any company or quantify financial losses, so it should be read as an indicator of preparedness rather than a measure of damage.

It is also worth noting that the survey was conducted by a vendor that sells workflow and security-related technology. Findings from vendor-led surveys are useful for spotting trends, but they reflect the respondents and questions chosen by the sponsor.

Background: what are AI agents?

AI agents, often called agentic AI, are software systems that can take actions on their own towards a goal, such as retrieving data, triggering workflows or interacting with other systems, rather than only generating text in response to a prompt. Because they can be given access to enterprise systems, their permissions and behaviour need to be managed much like those of an employee or a privileged account. Common governance measures include registering each agent, assigning an owner, testing it before launch and monitoring it once it is live. The report's "deployment gate" refers to an automated checkpoint that stops an agent from going into production until it passes defined security checks.

What to watch next

  • Whether more Indian organisations name a single accountable owner for agentic AI security.
  • Adoption of automated deployment gates and central registries for AI agents.
  • Progress in moving from automated detection to automated response, which the survey shows lags.
  • Any further surveys or regulatory guidance on AI governance and cyber resilience in India.

Frequently asked questions

What did the ServiceNow survey find about AI agent incidents in India?

The survey of 500 security leaders found that 92 per cent of Indian organisations had an AI agent security incident in the past year, and only 36 per cent of them contained it within 24 hours.

What is the name of the ServiceNow report?

The report is titled "Deployed but Ungoverned: The State of Agentic Security in India".

How many organisations have a named person accountable for agentic AI security?

Only 19 per cent have a single named individual accountable for agentic AI security governance, according to the report.

What share of firms found unregistered AI agents in production?

About 37 per cent of organisations discovered AI agents running in production in the last 12 months that were never registered or known to security teams.

Can Indian firms automatically respond to anomalous AI agent behaviour?

Mostly not. While 51 per cent have automated real-time detection, 67 per cent cannot automatically act when an agent behaves anomalously.

Own a stock in this story? Check it for red flags in seconds. Open the checker →

This article is for information only and is not investment advice.