Privileged access is the elevated permission that lets an identity change systems, reach sensitive data or override normal controls. In effect, it is the keys to the most valuable parts of an enterprise. For years, it was a contained problem, held by a small group of administrators and managed through periodic reviews. That world has disappeared. Privilege now lives in service accounts, cloud roles, data platforms and, increasingly, the AI agents that enterprises across Asia Pacific (APAC) are deploying at pace. Each is a holder of privilege in its own right, and together they are far more distributed, less visible and harder to govern than the static controls most organizations still rely on.
The scale of the shift is now visible in the market. IDC projects global security spending will pass $307 billion in 2026, with identity and access management software among the largest categories of that spend. Gartner, meanwhile, expects 40 percent of enterprise applications to embed task-specific AI agents by the end of this year, up from less than 5 percent in 2025, and every one of those agents is a privileged identity that has to be discovered, scoped and eventually retired. The consequence is already being priced in: Gartner predicts that by 2028, a quarter of enterprise breaches will trace back to AI agent abuse.
For CISOs across the region, this has moved from a horizon-scanning exercise to a governance gap opening in real time. Recent SailPoint research found that while 62 percent of security leaders say AI and automation are already creating new privileged accounts in their environments, only 18 percent rate themselves as highly prepared to govern that privilege. Even more concerning, recent data reveals that 97 percent of AI agents already have access to sensitive data, yet only 21 percent of organizations feel highly confident in managing AI agent security risks.
It is precisely in the distance between adoption and control that risk tends to concentrate. The instinct is to close that gap by buying more tooling, but that is the wrong lesson to draw. Most enterprises already run a mix of PAM, IGA and cloud entitlement tools, so tool presence has never really been the problem. The problem is posture, and that is where leaders should focus their attention.
Here are the priorities that matter most:
1. Redraw the map of what “privileged” actually means
Most privileged access programs were designed for a small, well-understood population of human administrators, and that definition simply no longer holds. Many organizations still classify only their people as privileged users, which leaves their fastest-growing risk surface effectively unmonitored. The first priority, therefore, is an honest inventory that captures not just who holds privilege, but what holds it, what that access can reach, and who is accountable for it. You cannot govern what you have never mapped.
2. Treat standing privilege as the liability it has become
The most dangerous pattern in AI adoption is the over-permissioned agent, granted broad and persistent access because the teams building it are optimizing for capability rather than constraint. Standing privilege that sits idle on a human account is already a risk; the same privilege attached to an autonomous, non-deterministic agent becomes a structural exposure. The answer is not a bigger vault, but a shift to Zero Standing Privilege, where access is granted just in time for a specific task and revoked the moment that task is complete. Organizations must also deploy specific defensive controls such as prompt security to redact sensitive data before it reaches large language models (LLMs) and centralized kill switches to instantly isolate rogue AI agents. When privilege expires by default rather than persisting indefinitely, the blast radius of a compromised agent or credential collapses, and the window an attacker can exploit closes with it.
3. Move from periodic reviews to continuous posture management
Quarterly access certifications made sense when environments changed slowly, but they cannot keep pace with agents that are created, cloned and connected daily. Leaders need to move from point-in-time review to continuous discovery and risk assessment, surfacing excessive or unused privilege as it accumulates rather than months later at audit. And because not all privilege carries the same exposure, the goal is not to review everything equally but to prioritize by risk, focusing on the access paths that reach sensitive or regulated data and on the identities that would do the most damage if abused.
4. Hold AI to the same standards of accountability as people
An AI agent that acts on the enterprise’s behalf should be as traceable as an employee who does the same work. In practice, that means every agent has a defined owner, scoped entitlements, a clear lifecycle and a full audit trail, not a reused human login and a set of permissions that nobody quite remembers granting. For enterprises operating under MAS guidelines in Singapore, PDPA obligations or the tightening AI governance expectations now emerging across the region, this is no longer simply good security hygiene. It is fast becoming a condition of doing business and, increasingly, a factor in how partners and procurement teams judge maturity.
The case for a unified approach
What ties these priorities together is a recognition that isolated controls can no longer keep up. In a dynamic digital landscape, siloed security tools inevitably leave critical gaps between them, and those gaps are exactly where over-permissioned agents and forgotten credentials go undetected. Investing in a unified and intelligent architecture helps eliminate these silos. Instead of adding one more product to an already crowded stack, it delivers a continuous and intelligent layer of protection woven across the entire enterprise, securing every identity, in every environment, at any speed. That unified lens is what makes it possible to govern human and non-human privilege with the same rigor, and continuously rather than in periodic bursts.
Govern first, and scale with confidence
None of this is an argument against AI. The productivity case is real, and the enterprises pulling ahead in APAC are the ones deploying agents aggressively. But the organizations that will still be ahead in a few years are those that build governance into the foundation rather than bolting it on after an incident, because privileged access has quietly become the discipline that determines whether AI adoption compounds into advantage or into liability. In the AI era, the real constraint on ambition is not how capable your agents are; it is how well you can account for what they are allowed to do.
Eric Kong is Managing Director, ASEAN at SailPoint. Eric leads the ASEAN business at SailPoint, a leader in identity security for the modern enterprise. He is responsible for driving strategy and GTM execution to deliver growth for the business and impact for customers. This in collaboration with the industry SailPoint’s growing partner ecosystem.
Prior to SailPoint, Eric held leadership roles across several global Information & Communications Technology (ICT) companies including Accenture, HP, Oracle, SAS, Vodafone and AWS. His experience across consulting, sales and strategic partnering has helped customers from across industries transform and grow in the Asia Pacific markets. Eric holds a Bachelor of Chemical Engineering degree from the National University of Singapore.
Editor’s note: This contributed article has been lightly edited for clarity, length, and style. Where appropriate, TNGlobal may verify, qualify or omit factual claims that cannot be independently corroborated. The views and arguments expressed remain those of the author.
Share your perspective: TNGlobal welcomes contributed insights and expert commentary from across Asia’s technology and innovation ecosystem. Submit a contribution for editorial consideration, or explore more conversations in our TNGlobal INSIDER and TNGlobal Q&A and Interviews archive.
Featured image: Riku Lu on Unsplash
Look beyond access management: Stay secure with a unified identity security approach