AI is actively transforming organizations across APAC, from Singaporean banks and Australian retailers to Japanese manufacturers and Indian startups. At the centre of this shift are AI agents: increasingly autonomous systems acting on behalf of people, teams, and entire companies.

But there’s a catch. Each AI agent needs deep access to systems and sensitive data to be useful, and those permissions are quickly becoming an open door for attackers. Many organizations across APAC still rely on outdated authentication methods, creating an absolute goldmine for threat actors.

Okta Threat Intelligence analysis reveals that this is leading to an alarming buildup of identity debt.

Outdated authentication elevates AI risks

Similar to technical debt, identity debt can accumulate when misconfigured permissions, orphaned accounts, and inconsistent security policies go unchecked. These gaps only increase the risk of breaches and complicate compliance efforts. They also add unnecessary operational overhead.

Agentic AI is now accelerating this problem at scale. When agents are granted broad permissions without frequent credential rotation, they can effectively become “super admins” with free run of an organisation. If compromised, threat actors gain the same privileges, effectively walking in and out of systems undetected.

Okta’s 2025 AI at Work report shows leaders globally know the risk. Executives ranked data privacy and security as their top two AI concerns by both severity and frequency. Though all eyes are on securing AI, only 10 percent of organizations have a well-developed strategy or roadmap for managing NHIs.

So while we’re speedrunning our way to agentic AI, these agents, if poorly managed, are becoming new attack vectors. All it takes is one prompt injection or persona switching attack to access sensitive access tokens that allow lateral movement.

With enterprises across APAC among those leading the adoption of no-code tools like Gemini Code Assist and platforms such as Zapier Central, IT teams are grappling with an explosion of user-generated agents that are fast, scalable, and often invisible.

Access controls are still a weak spot

Analysis of authentication methods used to connect security applications to Microsoft Copilot reveals a concerning reliance on insecure credential practices.

Basic authentication, used in 20 percent of cases, exposes accounts to attacks by requiring administrators to upload usernames and passwords without multi-factor protection. Meanwhile, 75 percent of integrations rely on static API keys, which, despite some improvements, are often stored in logs or plaintext files, giving attackers persistent access if intercepted. Enterprise-grade OAuth 2.0, offering short-lived tokens and scoped permissions, accounts for just 5 percent of connections, highlighting slow adoption of more secure methods.

As AI agents continue to scale, reliance on basic authentication and static tokens magnifies the potential impact of a single compromise. Without externalized access controls, these outdated approaches increase organizations’ blast radius, leaving sensitive data and critical systems increasingly vulnerable to breaches.

How to effectively manage the agent identity lifecycle

A different set of lenses is required to recognize key differences in provisioning, permissions, and visibility:

  • Programmatic provisioning: Agents are deployed from a CI/CD pipeline and need to be provisioned automatically without human interaction.
  • Highly-specific permissions: Each agent serves a use-case-specific purpose and requires granular-level permissions balanced with limited time access to minimize exposure.
  • Short life spans: Agents are dynamic and short-lived, which means rapid provisioning and de-provisioning are required to manage the field.
  • Visibility: Unlike human accounts, agents often lack traceable ownership and consistent logging; strong lifecycle and identity controls are required to track their actions and when they were taken.

The bottom line

Organizations must invest in secure authorization flows built for the AI environment, instead of bolting AI agents onto outdated authentication methods. Only by prioritizing identity-first strategies for AI agents can organizations stay secure and ahead of attackers.


Stephanie Barnett is VP and Interim GM, Asia Pacific & Japan at Okta. Based in Singapore, she leads the region’s operations and go-to-market initiatives to help organizations heighten their security postures through an integrated digital identity approach. Prior to Okta, she served in various leadership roles at Cloudflare, Salesforce, and IBM.

TNGlobal INSIDER publishes contributions relevant to entrepreneurship and innovation. You may submit your own original or published contributions subject to editorial discretion.

Featured image: Pexels

How the industry’s fear of change is costing printers customers they don’t even know they lost