JustUpdateOnline.com – The traditional concept of "shadow IT" is undergoing a radical and dangerous transformation. While IT departments were once focused on blocking unauthorized cloud storage or messaging apps, the emergence of "shadow AI" presents a far more complex challenge to corporate security. Unlike the static applications of the past, today’s AI tools function as autonomous agents that integrate deeply into corporate networks, creating a surge in unmanaged "machine identities."
Recent data underscores the financial and operational stakes. According to findings from IBM’s 2026 Cost of a Data Breach Report, 43 percent of organizations surveyed faced security incidents linked to shadow AI, more than doubling the 20 percent reported just a year prior. These specific breaches are costly, with an average price tag of approximately US$5.39 million per incident.
The Shift to Non-Human Identities (NHIs)
The primary concern is no longer just simple data leakage; it is the proliferation of Non-Human Identities (NHIs). When employees connect AI assistants to company calendars, CRM systems, or cloud repositories, they are often granting these tools persistent API keys and elevated permissions. These machine actors operate with the same level of access as human employees but frequently exist outside the view of security governance.
The Asia-Pacific (APAC) region has already seen significant fallout from these risks. In Japan, Panasonic Holdings revealed that personal data for 334,000 individuals was compromised when an employee used an unsanctioned generative AI platform. Similarly, in Singapore, food retailer Bee Cheng Hiang suffered a leak of 95,000 customer emails due to an unverified AI-driven mailing script. In Australia, a contractor’s use of ChatGPT exposed the sensitive health and personal records of 3,000 flood recovery applicants.
A Governance Crisis in the APAC Region
Traditional security models are often built around human users and defined perimeters, but AI agents bypass these assumptions. Takanori Nishiyama, Senior Vice President and APAC & Japan Country Manager at Keeper Security, notes that an AI agent with access to a production database carries the same risks as a high-level human administrator, yet it functions at machine speed.
The "Identity Security at Machine Speed" report highlights a significant oversight gap. Among 3,200 IT leaders surveyed globally, 53 percent of those in the APAC region identified the management of AI-related NHIs as a primary weakness in their security posture—a figure 10 percent higher than the global average. Furthermore, nearly half of regional respondents admitted they lack visibility into the specific AI tools their employees are utilizing, and only 38 percent have fully implemented Privileged Access Management (PAM) solutions.
Evolving Regulatory Frameworks
Governments across Asia-Pacific are beginning to address the accountability of AI agents. In Japan, the Ministry of Economy, Trade and Industry (METI) recently updated its AI Guidelines for Business to include specific definitions for AI agents.
Singapore has taken a proactive stance, with the Infocomm Media Development Authority (IMDA) launching a governance framework for agentic AI that insists every digital agent be linked to a responsible human party. The Cyber Security Agency of Singapore (CSA) has also released specific guidelines on securing these technologies. Meanwhile, in Australia, the National AI Centre and the Office of the Australian Information Commissioner (OAIC) are urging organizations to maintain human oversight and avoid inputting sensitive personal data into public AI models.
Strategies for Modern AI Governance
Security experts argue that simply banning AI is an ineffective strategy, as employees will likely find workarounds that further hide the technology from IT oversight. Instead, a more robust approach involves extending identity-centric security to both humans and machines.
Key recommendations for organizations include:
- Continuous Discovery: Security teams must actively identify all AI tools and integrations running within their environment to uncover hidden NHIs.
- Real-Time Monitoring: Recording and auditing AI-driven sessions allows for immediate oversight of automated workflows.
- Zero Standing Privilege: Organizations should aim to grant AI agents elevated access only when necessary for a specific task, revoking it automatically once the job is complete.
- Lifecycle Management: Machine identities should be subject to the same rigorous authentication, credential rotation, and deprovisioning protocols as human accounts.
By adopting a "Zero Trust" framework—where every access request is verified regardless of whether it originates from a human or a machine—businesses can leverage the benefits of AI while maintaining control over their most sensitive digital assets.
