India, Oct. 4 -- Indian enterprises appear to have moved quickly to put rules around Artificial Intelligence (AI). The harder part is making those rules hold when AI agents actually start working across enterprise systems.

A new Delinea study presents an unusual contradiction. 99% of surveyed Indian organisations say they have a formal policy governing what data AI tools and agents can access, and 87% say those policies are actively enforced. Yet 84% reported that an AI tool or agent accessed sensitive data beyond its intended scope during the past year.

The finding suggests that the next phase of enterprise AI security may be less about writing governance policies and more about enforcing access boundaries as AI systems act across data...