Cloud services, automation and artificial intelligence are reshaping corporate IT environments. But as their use grows, so does the number of digital identities that do not belong to people. These so-called non-human identities are emerging as a rapidly growing source of risk beyond traditional cybersecurity . A new KPMG thought leadership (TL) explores why businesses need to start paying closer attention.
When an employee joins a company, they receive a user account and access to the systems they need. As their role changes, their permissions can be adjusted, and when they leave, their access is removed. But in today's digital environments, people are no longer the only ones interacting with systems. Applications, services, automated processes and AI agents account for a growing share of these interactions, and they need digital identities of their own.
What are non-human identities?
Non-human identities (NHIs) are digital identities and credentials that allow machines, applications, services, scripts and automated processes to authenticate and communicate with one another without direct human involvement. They include service accounts that run background processes, API keys and tokens that connect different services, RPA bots and technologies controlling physical devices and processes. Increasingly, they also include autonomous AI agents.
These identities play a vital role in modern organisations. They allow applications and services to connect to other systems, exchange data and carry out tasks automatically across cloud environments, software platforms and internal IT infrastructure.
The issue is not the existence of these identities themselves. The challenge lies in their growing numbers, the level of access they may have, and limited visibility over where they exist and who is responsible for them.
Digital identities are growing faster than businesses can manage them
The rise of cloud services, automation and, more recently, agentic AI is accelerating the creation of new non-human identities. Some studies suggest that as early as 2025, in many enterprise environments there are already around 80 non-human identities for every human identity, with some environments reporting ratios as high as 144 to 1.
As their numbers grow, so does the complexity of managing them. Unlike human identities, which businesses typically manage through established processes, NHIs are often created programmatically and can fall outside standard identity management practices. This can result in unused identities, unclear ownership or poorly managed access lifecycles.
NHIs may also have extensive permissions. Risks highlighted in the TL include passwords and other sensitive credentials embedded directly in code, credentials not rotated frequently enough, and tokens with overly broad permissions. Some security measures commonly used for human identities, such as multi-factor authentication, also cannot be applied to NHIs.
If one of these identities is compromised, the consequences can extend far beyond the IT department. Compromised NHIs can disrupt payments, take customer-facing services offline or break integrations that critical business processes depend on.
AI agents add a new dimension to the challenge
The rise of autonomous AI agents deserves particular attention. These systems can independently carry out tasks, retrieve data and communicate with external APIs. Unlike traditional automated processes, they can dynamically combine multiple actions across different systems without requiring a person to initiate every individual step.
This also brings new security concerns. One example is prompt injection, where malicious input can influence the behaviour of an AI agent and potentially lead to the misuse of its permissions. Other risks include movement across systems and the creation of additional identities that may go undetected.
The more tasks organisations automate, the more important it becomes to understand not only who is accessing corporate systems and data, but also what is accessing them.
Author
Alexander Zagnetko
KPMG Global AI Initiative Coordinator
How can businesses bring non-human identities under control?
Effective management of non-human identities can be built around four key steps:
Measuring progress is just as important. Relevant indicators include the proportion of NHIs with a clearly identified owner, credential lifespans, the speed at which exposed sensitive credentials are detected and remediated, and the number of inactive or orphaned identities.
Identity management is no longer just about people
The challenges associated with AI agents are not limited to those created by an organization and its employees. Agents from the external ecosystem may not be malicious, yet when accessing organizational infrastructure and assets, they can perform unauthorized actions, causing problems ranging from data leakage to overloading computing resources This makes it crucial for organisations to approach their security systematically and give them the same level of attention they already devote to human identities.
For businesses, it is therefore no longer enough to ask who has access to our systems? It is becoming just as important to understand what is accessing them, what permissions that identity has, and who is responsible for it.
KPMG Cyber Security: Non-Human Identities (NHI) Management – From Risk to Resilience
Explore the risks associated with non-human identities and how organisations can manage them in the new KPMG study.
Contact us
Should you wish more information on how we can help your business or to arrange a meeting for personal presentation of our services, please contact us.