AI access policies fail at the point of enforcement in Indian enterprises

Widespread data overreach, persistent credentials and weak access traceability show that governance frameworks are not consistently controlling what AI agents can do with sensitive information.

Indian enterprises have widely adopted policies governing how AI tools and agents access corporate data, but those policies are not consistently preventing AI systems from reaching sensitive information beyond their intended scope.

Delinea’s 2026 Identity Security Report: The AI Enforcement Gap found that 84 percent of Indian organisations experienced an incident in the past year where an AI tool or agent accessed sensitive data beyond its intended scope, despite 99 percent reporting that they have a formal policy governing AI access to data.

The findings expose a growing gap between governance and control as enterprises give AI systems access to employee information, customer data, financial records, source code and other sensitive resources.

The findings suggest enterprises are moving beyond AI governance policies and focusing more on controlling what AI agents can access, how long that access remains active and who remains accountable for their actions. This creates opportunities for identity security providers, MSSPs, DevSecOps specialists and AI implementation partners.

Indian organisations are ahead of their global peers on several AI governance measures.

The report found that 87 percent actively enforce AI data-access policies, compared with 71 percent globally. Yet Indian enterprises also give AI broader access to sensitive information, including employee records, customer data, financial information and source code.

“India’s enterprises have done the hard work on AI governance. Almost every organisation has a policy, and most enforce it more rigorously than their global peers,” said Cynthia Lee, vice president, APJ, Delinea.

“But AI agents here also reach more customers, employees and financial data than the global average. At that point, a policy alone stops being enough.”

The study also found that 98 percent of organisations believe they could demonstrate compliant AI access to a regulator, despite the widespread overreach reported during the past year.

Indian enterprises are detecting AI overreach faster than their global peers. Around 34 percent identified their most recent scope violation as it occurred, compared with 20 percent globally. Nearly half could immediately revoke both an AI tool’s credentials and an active agent session.

However, the findings suggest that detection and response capabilities are advancing faster than the controls governing how AI systems receive and maintain access in the first place.

Persistent credentials weaken accountability

One of the clearest weaknesses appears in the way organisations manage AI agent credentials.

While almost every respondent said they treat AI credentials such as API tokens and Model Context Protocol configuration files as privileged credentials, 45 percent reported that some credentials remain active until the next audit, even after the task they were issued for has ended.

This creates standing access that can allow AI agents to retain permissions they no longer require.

The report also highlights a gap in accountability. Nearly every organisation requires a named individual to approve access before an AI system reaches a new sensitive data source. However, only 47 percent said they can always trace a sensitive AI access event back to that individual.

Software development environments remain another weak point. Only 43 percent of Indian organisations can enforce AI access controls at the point of action within CI/CD pipelines, even as coding agents gain access to source code, development environments and cloud infrastructure.

According to Delinea, enterprises need controls that extend beyond policy documents and login authentication. Access decisions must be enforced while AI agents are performing tasks, with visibility into who approved the access, what actions were taken and whether those permissions remained valid.

While governance frameworks are now largely in place, organisations are likely to focus more on privileged-access management, short-lived credentials, runtime authorisation and AI activity monitoring as they look to close the gap between policy and enforcement.