Home / Posts / SpyCloud 2026 NHI Report

SpyCloud 2026 Report: The Silent Threat of Non-Human Identity Exposure

March 20, 2026 Dillip Chowdary

While the industry has spent decades securing human credentials through MFA and biometrics, a new and far more dangerous frontier has emerged: Non-Human Identities (NHIs). SpyCloud's 2026 Non-Human Identity Exposure Report reveals a staggering 314% increase in leaked API keys, service accounts, and OAuth tokens over the past 12 months. As organizations embrace agentic AI, Kubernetes, and serverless microservices, the ratio of NHIs to human identities has reached an alarming 45:1, creating a massive, unmanaged attack surface that traditional tools are blind to.

The NHI Anatomy: The "Machine Credentials" Problem

NHIs are the "glue" of modern digital infrastructure. They include CI/CD pipeline secrets, cloud infrastructure identities (IAM roles), database connection strings, and bot-to-bot authentication tokens. Unlike human users, NHIs don't have "working hours," they don't use MFA, and they often possess excessive, static privileges that allow them to bypass traditional perimeter defenses.

The report highlights that hardcoded secrets in public and private repositories remain the primary source of exposure. SpyCloud's researchers identified over 12.7 million active secrets across GitHub, GitLab, and Bitbucket in 2025 alone. These aren't just old, discarded keys; 68% of identified secrets were still valid at the time of discovery, providing attackers with immediate, high-level access to sensitive PostgreSQL databases and production AWS/Azure environments. The shift to Infrastructure-As-Code (IaC) has inadvertently made it easier for developers to accidentally commit secrets into version control.

Technical Benchmark

The average Mean Time to Detection (MTTD) for a compromised NHI is 142 days, compared to just 11 days for a human credential compromise. By then, an attacker has typically achieved full tenant persistence.

Shadow AI and Token Proliferation: The New Attack Vector

The explosion of Generative AI has added fuel to the fire. Developers and business units are increasingly deploying Shadow AI—unauthorized AI agents that require API access to various SaaS platforms like Salesforce, Slack, and Jira. These agents often store bearer tokens in insecure browser caches, unencrypted local files, or environment variables that are exposed in container logs. SpyCloud found that 42% of NHI exposures in Q4 2025 were directly linked to AI orchestration tools and experimental LLM integrations.

Furthermore, the report introduces the concept of Token Chaining. Attackers are now using a single compromised low-level token (e.g., a Slack bot token) to "pivot" and extract more powerful credentials (e.g., an AWS Access Key) by monitoring internal communications or exploiting over-scoped permissions in integrated apps. This lateral movement is often invisible to traditional Identity and Access Management (IAM) tools because the behavior mimics "normal" machine-to-machine traffic.

Secret Sprawl in the Cloud: The Serverless Risk

The report also dives into the risks of Serverless functions (Lambda, Google Cloud Functions). Because these functions are ephemeral, developers often inject secrets as Environment Variables. However, if the function's logs are not properly sanitized, these secrets are leaked to CloudWatch or Stackdriver, where they can be harvested by anyone with read-access to the logging service. SpyCloud identifies this as the fastest-growing category of NHI exposure, with a 200% year-over-year growth.

Strategic Recommendations for 2026: Machine Identity Governance

"Organizations must treat NHIs with the same rigor as human identities," says SpyCloud CTO Trevor Hilligoss. "The era of 'set and forget' service accounts is over. We need automated secret rotation, just-in-time (JIT) provisioning for machines, and identity-centric observability that can detect anomalous machine behavior in real-time."

SpyCloud recommends a Zero Trust Architecture for Machines. This includes moving away from static API keys toward short-lived, ephemeral credentials (e.g., using HashiCorp Vault or AWS Secrets Manager with auto-rotation) and implementing Strict Scoping (Least Privilege) for all service accounts. As we move further into the age of autonomous agents, the ability to manage and secure Non-Human Identities will be the deciding factor in an organization's cyber resilience.

Protect Your Development Environments

Prevent accidental exposure of sensitive data in your logs and test environments. Use Data Masking Tool to sanitize your data before it leaves your secure perimeter.

Try Data Masking Tool →