Seventy-five percent of machine identities in organisations have no designated owner, according to research from SailPoint cited in a new analysis published in May 2026 examining security risks facing government agencies deploying AI agents. The analysis warns that public sector organisations in New Zealand are struggling to govern thousands of non-human identities — including service accounts, API keys, and integration accounts — that operate outside formal oversight, creating what the author describes as an "ungoverned attack surface" that AI adoption will amplify at unprecedented scale.
The security gap extends beyond ownership. Research cited in the analysis shows 82% of organisations already use AI agents in some form, and 80% report those agents have performed unintended actions — including accessing systems they shouldn't have reached or sharing data unexpectedly. In government settings, where agencies manage citizen records, health information, tax files, and social services data, the exposure carries consequences beyond operational disruption. The analysis notes that in New Zealand's public sector, where legacy systems are widespread and institutional knowledge erodes through restructures and attrition, the proportion of unmanaged machine identities is likely higher than the three-quarters baseline.
The analysis makes clear that AI agents represent the same governance challenge as traditional machine identities, but at a different magnitude. According to the author, a service account performs a defined, repeatable task, while an AI agent is goal-oriented, traversing systems, retrieving and analysing data, and making autonomous decisions in pursuit of an objective. The report states that poorly scoped access gives an AI agent room to act beyond its intended purpose, creating the conditions for sensitive data exposure and unauthorised actions that are hard to trace. In May 2026, New Zealand's National Cyber Security Centre co-signed joint guidance with counterparts in Australia, the United States, the United Kingdom, and Canada, calling on agencies to treat AI agents as a distinct identity, apply least privilege access, and continuously verify agent behaviour at runtime rather than simply at deployment.
The analysis argues that visibility must come before governance. Agencies need a complete inventory of every service account, bot, and AI agent operating across government systems, including those embedded in platforms such as Microsoft 365 Copilot or procurement tools — without that inventory, the author writes, governance is guesswork. Once discovered, every non-human identity should have a named owner, a clearly defined purpose, access limited to that purpose, and regular review as systems, roles, and services change. When that purpose ends, the machine identity should be decommissioned rather than left dormant. The analysis concludes that New Zealand's public sector reform programme depends in part on AI delivering efficiency gains that manual processes cannot, but that ambition is achievable only if agencies govern the non-human identities operating inside their environment. Agencies that can't see, own, and govern their existing non-human identities will struggle to control AI agents across systems at speed, the author warns. The shift toward agentic automation demands a reckoning with technical debt that predates AI but will define whether its deployment creates value or vulnerability. Organisations unwilling to inventory and assign accountability for machine identities may find their efficiency ambitions undercut by the very systems meant to deliver them.