From Agentic AI Boom to API Doom? Securing the Foundation of APAC’s AI Ambitions 

Agentic AI (Artificial Intelligence) has been unleashed across the Asia Pacific, and it’s learning, acting, and deciding at machine speed. According to an April 2025 Microsoft report, APAC is leading the world with 53% of business leaders reporting they have fully automated workstreams or entire business processes using AI agents to solve real business problems.

For enterprises, the promise of a hyper-efficient future and automation solving talent gaps and population challenges, broad adoption of intelligent applications giving predictive insights, and tying it together to gain an unbeatable competitive edge. This is the engine of our next great transformation, backed by a projected USD $110 billion in regional AI spending by 2028, according to IDC.

The problem is that this entire AI-powered future is being built on the most brittle part of our digital foundation: our APIs.

APIs have evolved from simple connectors into the execution layer for the enterprise. They are the arteries through which these new semi-autonomous AI agents carry out their instructions in the real world. If those arteries are left unprotected, every system, every piece of data, and every business process they touch becomes critically vulnerable.

In the rush to innovate, APIs are being deployed without disciplined governance, consistent security controls, or a clear line of sight. This gap between our AI ambitions and our security execution is widening by the day. And unless we make API security a boardroom-level priority, the region’s AI boom may unfortunately be headed to API doom.

Awareness Is High, Execution Is Not 

The adoption rate of generative AI is staggering. Beyond the Microsoft study above, a recent IDC study reveals that 42% of organisations in Southeast Asia have implemented agentic AI, with that number expected to hit a massive 86% by mid-2026. The agents are already here and multiplying fast.

Ironically, we know we have a problem. According to F5’s latest survey and report, 2025 Strategic imperatives: Securing APIs for the age of Agentic AI in APAC, nearly two-thirds (63%) of organizations say API security is “very important.” Awareness isn’t the issue. Execution is.

The same F5 report reveals that only 42% of enterprises have a mature API governance program, and a mere 22% have a dedicated API security function. This is a massive disconnect. It’s like knowing your car’s brakes are faulty but deciding to accelerate anyway.

The New Threats Are Old Problems, Amplified by AI 

So, what does this failure to execute look like on the ground? According to IDC, the average enterprise investing in Generative AI increases their API count by roughly 5x. APIs are how you build AI models, how you use them with techniques like Retrieval Augmented Generation (RAG), and how you attack them via patterns like model inversion, prompt injection, and best-of-n jailbreaking.

In other words, AI applications are APIs all the way down, and way too many enterprises are doubling down on expanding an attack surface they already don’t understand, and opening themselves up to new classes to attack via these new interfaces.

Layer in the speed of automation and non-deterministic nature of AI, and AI agents making thousands of autonomous calls per minute can turn simple misconfigurations into catastrophic failures and exploits. A minor oversight can spiral into a large-scale data breach or service outage before a human can even react.

Then you have the ghosts in the machine of Shadow and Zombie APIs. Shadow APIs are unmanaged, undocumented endpoints that developers create for all kinds of reasons including convenience. They’re unmarked doors to your digital castle. While 36% of enterprises rate them as a high risk, our data shows only 38% believe they can effectively detect them.

Zombie APIs are even more sinister. These are the forgotten endpoints from old application versions that were never properly decommissioned. A staggering 15% of organisations admit they have minimal ability to find them.

Much like the bots and scripts we defend against today, autonomous agents don’t just use the systems you approve; they discover and interact with the neglected ones you can’t see. If your defenses are still tuned for predictable, human-driven traffic, you’re fighting the last war while the next one is already happening.

Why Your Old Defences Will Fail 

Despite these clear and present dangers, most organisations are still leaning on perimeter-based tools like web application firewalls (WAFs) and traditional identity and access management (IAM). While still necessary, these defences were designed for different problems in a different era. They are gatekeepers built to manage human-scale interactions, not the relentless, high-velocity traffic of autonomous bots and AI agents.

Even our security leaders know they’re behind. A recent Salesforce study found that 57% of IT security leaders in APAC are not fully confident they have the appropriate guardrails to deploy AI agents securely.

From Reactive to Resilient: Five Boardroom Imperatives 

The good news is that change is coming. Sixty-nine percent of enterprises plan to increase their API security spending. But throwing money at fragmented tools will only reinforce the silos we need to break down.

To build a resilient foundation for AI, organisations need a unified, lifecycle approach to API security. This requires five non-negotiables:

1. Assign C-Level Ownership. This isn’t just an IT problem; it’s a core business risk. API governance must be owned at the executive level and aligned with the enterprise AI strategy. By default this falls to the CISO in most cases, but CISOs often want this responsibility to live with Application Development or DevOps teams. If you don’t have a clear owner, you won’t have a clear plan.
2. Prioritise Lifecycle Controls. Security can’t be a final checkpoint. It must be a guardrail embedded across the entire lifecycle—from discovery and posture management to runtime protection and testing.
3. Embed Agent-Aware Observability. You can’t protect what you can’t see. You need real-time visibility that understands and flags anomalous behaviour from autonomous agents, not just humans.
4. Enforce Zero-Trust Policies Consistently. The fundamentals matter. Apply robust, OWASP-based policies to every API, for every user, whether it’s a person or a machine.
5. Link API Security to Business Outcomes. Translate security metrics into business risk. A strong governance framework demonstrates how a secure API posture is an enabler of faster, safer innovation.

Secure the API Connection, Secure the Future 

Ultimately, trust in AI is inseparable from trust in the APIs that serve it.

Every interaction, whether initiated by a human or an AI agent, must be authenticated, authorised, and continuously monitored within a framework of unified governance.

The choice for leaders across APAC is stark. To truly unlock the immense promise of agentic AI, we must first secure the connections that we need to make it work. In the end, our AI future is only as strong as the connections we use to hold it together.