Indian firms struggle to secure rapidly expanding AI agents
A new study reveals that while Indian companies are quickly adopting AI agents, they lack the governance and automated tools to stop security incidents in real-time.

The short version
- 92% of Indian organizations reported an AI agent security incident in the past year.
- Only 36% of those affected were able to contain the security incident within a 24-hour window.
- 37% of companies discovered 'Shadow AI'—agents running in production over the last year that were never registered with security teams.
- Only 18% of surveyed organizations have a unified, single-interface view of their AI agents.
- 83% of organizations have AI agents interacting with physical environments or operational technology.
Why it matters
As companies move from simple chatbots to autonomous agents that can perform actions, the risk of unmonitored behavior increases. For the average user, this means that the systems handling data and physical infrastructure may lack the necessary 'brakes' to stop errors or attacks quickly.
The gap in agentic security
Indian enterprises are increasingly integrating AI agents into their daily operations, but the security infrastructure required to manage these tools is lagging behind. A study by ServiceNow, which surveyed 500 security leaders in the country, shows a significant disconnect between adoption and oversight. While many leaders believe their security is established, the actual controls in place often do not match that confidence.
The data reveals a stark reality regarding incident management. Although the vast majority of organizations encountered security issues involving AI agents last year, very few were able to resolve them quickly. This suggests that while companies are eager to use the technology, they are not yet prepared for the speed at which these autonomous systems can operate or the complexity of their risks.
The report, titled Deployed but Ungoverned: The State of Agentic Security in India, highlights that the current pace of deployment is outstripping the creation of governance frameworks. Without proper oversight, these agents can operate in a vacuum, potentially creating vulnerabilities that security teams are not equipped to see or manage effectively.
Visibility and accountability issues
One of the primary hurdles for Indian security teams is a lack of basic visibility into what is actually running on their networks. Only 18% of the surveyed organizations have a unified, single-interface view of their AI agents. This lack of a central dashboard makes it difficult for IT departments to track the lifecycle and behavior of various autonomous tools across different departments.
Accountability is also a major missing piece in the current landscape. The study found that only 19% of organizations have a single named individual responsible for AI agent security governance. When no one is specifically tasked with overseeing these systems, it becomes much harder to enforce consistent policies or ensure that security standards are being met across the entire company.
This lack of oversight leads to the rise of 'Shadow AI,' where systems are deployed without the knowledge of the security team. About 37% of organizations discovered agents active in live environments during the last year that were never logged with security staff. These hidden systems become blind spots, making it nearly impossible for security professionals to protect the company's data or infrastructure from potential threats.
Detection vs. automated response
The research indicates that while Indian organizations are getting better at spotting suspicious behavior, they are still struggling to stop it automatically. About 51% of organizations have implemented automated real-time detection for anomalous behavior in their AI agents. However, 67% of those same organizations cannot actually take automatic action when an agent begins to behave in an unusual or dangerous way.
This creates a dangerous delay between identifying a threat and containing it. Because AI agents can operate at machine speed and continuously, a human-in-the-loop response may be too slow to prevent significant damage. The study suggests that the current model of detection without immediate remediation leaves a wide window of opportunity for security breaches to escalate.
The role of security teams is often being sidelined during the early stages of development. Only 46% of organizations involve security teams before AI agent development begins. When security is treated as an afterthought rather than a foundational requirement, it is much harder to build in the automated controls necessary to handle the unique risks posed by agentic workflows.
Risks to physical environments
The reach of AI agents in India extends far beyond standard office software and into physical infrastructure. The study found that 83% of organizations have AI agents operating in, or interacting with, operational technology (OT) and physical environments. This means AI is increasingly involved in the systems that manage power, manufacturing, and other critical physical assets.
Despite this widespread integration, the governance for these specific settings is severely lacking. Only 29% of organizations have a separate governance framework designed specifically for these physical and operational environments. This gap is particularly concerning because a security failure in an OT environment can have real-world consequences that go beyond just data loss.
To address these risks, experts are calling for a more structured approach to agent management. Kulmeet Bawa, GVP and MD for India and SAARC at ServiceNow, stated, "You can’t secure or govern what you can’t see." He advocates for a system where every agent is registered, assigned a specific owner, and made visible in one place with automatically enforced controls to ensure safety and compliance.
The path to better security
The data suggests that organizations using unified platforms for agentic security are seeing better results than those trying to manage systems piecemeal. Among respondents using such platforms, 62% had automated real-time alerts for unusual behavior, compared to 43% of those who did not. These platforms also showed a higher rate of automated remediation for policy violations, at 45% versus 24% for non-users.
These findings come at a time when Indian regulators are beginning to take a more active role in oversight. CERT-In issued an advisory in October 2026 focusing on the specific risks posed by AI agents. Additionally, existing directions from 2022 require companies to report security breaches within a six-hour window of discovery, putting more pressure on companies to improve their detection capabilities.
The ultimate goal for Indian enterprises is to move toward a model where security is built into the deployment gate. Currently, only 30% of organizations have an automated gate that blocks AI agents until they clear security checks. By improving visibility and automating the response to threats, companies can begin to bridge the gap between the rapid adoption of AI and the need for reliable, secure governance.
What to watch
- How Indian regulators will enforce the October 2026 advisory on AI agents.
- The adoption rate of automated deployment gates to block unverified AI agents.
- The development of specific governance frameworks for AI in operational technology (OT).
Try it
Check if your company has a 'Shadow AI' policy or a registry for which AI tools are currently authorized for use in production.
Sources and links
- ReportingCNBCTV18: the article we sourced
First reported by CNBCTV18. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.