AI in the Enterprise: Security Struggles to Keep Up with Agentic AI

As businesses race to integrate AI into their daily operations, a growing shadow looms over cybersecurity. The latest findings from the 2026 SANS State of Identity Threats & Defenses Survey highlight that organizations may be falling behind in securing the very systems meant to accelerate their workflows. The surge in non-human identities (NHIs)—ranging from service accounts and API keys to automation bots—is colliding with the rapid adoption of agentic AI, creating a perfect storm for security gaps.

The survey, based on interviews with over 500 global security professionals, reveals that 76% of organizations report an increase in NHIs. Even more striking, 74% are already using AI agents or automated systems requiring credentials. The resulting expansion has caused NHIs to double or even triple within organizations, quietly operating behind the scenes. While traditional NHIs follow predictable patterns, agentic AI interprets instructions and acts autonomously, often with privileged access to critical infrastructure and sensitive data. This unpredictability raises risks akin to giving an over‑privileged insider superhuman speed and discretion. The threat of AI “hallucinations”—unexpected, erroneous actions—adds another layer of concern.

Forrester has warned that a publicly reported breach caused by agentic AI is likely by the end of 2026, urging companies to implement a “minimum viable security” framework for these deployments. Yet, SANS Institute’s findings suggest that most organizations are far from prepared. The study reports that 92% fail to rotate machine credentials within the recommended 90-day window, while 59% rotate less than half of their NHIs quarterly, and 15% don’t even know their rotation schedule. Alarmingly, 5% of respondents are uncertain whether agentic AI is even running in their environments. Manual access reviews and ticket-based provisioning, long relied upon, simply cannot scale to handle the growing volume and speed of NHIs across DevOps, cloud, and SaaS systems.

Richard Greene, certified instructor at SANS Institute, warns that enterprises are giving AI operational authority faster than they can build governance structures to manage it. Some early signs of progress exist—nearly 40% of organizations now include human-in-the-loop approvals for AI agent actions—but the real test lies in managing agentic AI as it moves from pilot projects to core operations. To mitigate these risks, SANS recommends adopting secrets vaults, automated credential rotation, and strictly scoped least-privilege access, while emphasizing that security efforts must scale alongside the rapid growth of NHIs.

What Undercode Say:

The SANS report uncovers a critical tension in enterprise cybersecurity: the speed of AI adoption is outpacing traditional security practices. Non-human identities, once a manageable toolset for automation, have become a sprawling risk as agentic AI enters the mix. The unpredictability of AI actions—essentially machine-driven decision-making with elevated privileges—turns a classic insider threat into a new, machine-accelerated paradigm. Unlike human staff, AI can simultaneously interpret multiple instructions and perform actions that weren’t explicitly anticipated, amplifying both risk and potential damage.

Credential hygiene is emerging as the Achilles’ heel of AI security. Organizations that fear breaking service accounts by rotating credentials are inadvertently creating long-lived attack vectors. A single compromised AI agent could pivot across critical systems at unprecedented speed, exposing sensitive data and potentially triggering regulatory repercussions. Current governance approaches—manual access reviews, ad hoc ticketing, and periodic rotation—fail to scale in high-velocity AI environments. The study highlights the necessity of automated solutions, yet adoption remains inconsistent.

Another dimension is organizational awareness. With 5% of respondents unaware of AI deployments in their own environment, the gap between technology rollout and oversight is stark. This indicates that AI risk is often invisible until an incident occurs, a concern that traditional risk management frameworks may not account for. Moreover, the lack of robust human-in-the-loop integration leaves systems prone to decision drift and unintended behaviors.

The human factor also remains critical. Even the best AI governance mechanisms rely on proper configuration, oversight, and continuous auditing. Without these, AI systems behave like rogue actors operating at machine speed, challenging traditional security paradigms. Enterprises must adopt a dual strategy: invest in AI-specific security controls and simultaneously build AI governance into organizational culture. Secrets vaults, automated rotation, and least-privilege enforcement are foundational, but the bigger challenge is continuous scaling as AI agents proliferate.

Finally, the report hints at a broader industry trend: AI adoption is no longer a question of efficiency—it is a question of resilience. Enterprises that treat AI as a tool rather than a new class of identity are likely to encounter breaches sooner rather than later. Predictive monitoring, real-time auditing, and adaptive credential management should become standard practices for organizations embracing agentic AI. The SANS survey essentially signals that AI security is moving from advisory to urgent operational necessity.

Fact Checker Results:

✅ 76% of organizations report growth in NHIs – aligns with SANS survey data.
✅ 74% of organizations already use agentic AI requiring credentials – confirmed by report.
❌ Only 5% unaware of AI deployments – might underestimate true unawareness due to survey sample bias.

Prediction:

The rapid adoption of agentic AI combined with poor credential hygiene suggests that 2026 will see the first high-profile AI-related data breach in enterprise environments. Companies that fail to scale automated governance and implement secrets management will be most at risk. Regulatory scrutiny is likely to increase, forcing organizations to adopt real-time monitoring and AI-aware cybersecurity frameworks. ⚠️

If you want, I can also create a visual diagram showing how NHIs and agentic AI interact within an organization, making the security risk easier to grasp. Do you want me to do that?

🕵️‍📝✔️Let’s dive deep and fact‑check.

References:

Reported By: www.infosecurity-magazine.com
Extra Source Hub (Possible Sources for article):
https://www.stackexchange.com
Wikipedia
OpenAi & Undercode AI