Key Takeaways
Gartner found 63% of organizations have fully or partially implemented zero trust, yet most programs cover half or less of the environment.
Six common implementation gaps can leave organizations vulnerable: product thinking, one-time verification, standing privilege, policy debt, missing visibility, and cloud drift.
Attackers go after what zero trust leaves out, so uncovered areas carry most of the risk.
Conditional access needs re-checking during a session, and privileged access should expire on its own.
The NSA’s 2026 guidelines break implementation into a Discovery phase, Phase One, and Phase Two.
Seeing what approved users and devices do after sign-in is the step most programs skip.
Ask a security team whether they’ve implemented zero trust and the answer is usually yes. Ask which parts of the environment it actually covers, and the answer takes longer.
That pause is what this blog is about. The identity provider is in place, MFA prompts show up on time, and the program has its own slide in the board deck. Then someone signs in with a stolen but valid account and spends weeks moving between servers that were never in scope. Nothing fires, because nothing was looking there.
As you read through the six gaps below, keep one question in mind: for each one, who would find it first in your environment, your team or an attacker?
What Zero Trust Is Supposed to Do
Zero trust drops the assumption that anything inside the network is safe. A laptop on office Wi-Fi earns no more trust than one in a hotel lobby. The model rests on three principles: verify every request using identity and context, give users and workloads only the access a task needs, and design as if an attacker is already inside.
The urgency comes from how far the perimeter has dissolved. NIST’s 2025 practice guide tested its example builds against an environment with several cloud platforms, a branch office, and employees working from coffee shop Wi-Fi. One firewall at headquarters can’t stand in front of all of that.
Enterprises have responded. In Gartner’s 2024[1] State of Zero-Trust Adoption Survey of 303 security leaders, 63% said they had fully or partially implemented a zero-trust strategy, and the model keeps spreading.
In January 2026, Gartner predicted that half of all organizations will apply zero trust to data governance by 2028[2], as unverified AI-generated data piles up. One older detail still deserves attention, though. In that same adoption survey, 56% said their main reason was that zero trust is considered industry best practice. A program started because everyone else is doing it tends to get scoped around what’s easy to deploy, rather than what’s most at risk.
What Are the Main Challenges in Adopting a Zero Trust Security Model?
Gartner’s less flattering finding is about coverage. For most organizations, zero trust addresses half or less of the environment and mitigates a quarter or less of enterprise risk. Only 16% expected to reach 75% coverage.
So why do programs stall so early?
ChallengeWhat it does to the program
ComplexityNIST’s Alper Kerman, co-author of its 2025 guide[3], noted that every network is different, so each architecture is a custom build, and the experts who can do it are scarceLegacy systemsMainframes and in-house applications often can’t support modern authentication, so they get exemptedStaffing41% of organizations expected staffing needs to riseBusiness frictionZero trust touches every team, and support fades when people feel slowed down without seeing resultsStarting pointPrograms either begin too broad to finish or too narrow to grow
Even government guidance has had to adjust. When CISA released version 2.0 of its Zero Trust Maturity Model, it added an “Initial” stage between Traditional and Advanced because many organizations found that jump too big to make in one go.
These are the causes. The gaps are what they leave behind.
Dynamic threat monitoring
Proactive Cyber Defense
Zero Trust in Action
The Zero Trust Implementation Gaps That Leave Enterprises Exposed
None of the six below is exotic. Most teams will recognize at least two of them in their own environment.
Gap 1: Zero trust bought as a product
It often starts with a purchase. A tool arrives, described as zero-trust-ready, MFA goes live, and the project turns green.
What the organization has is a set of controls deciding in isolation. The identity provider judges sign-in, the endpoint agent watches the device, and the firewall enforces its rules, but none of them shares what it knows. An attacker who replays a stolen session token may avoid an MFA prompt altogether, especially where session protection and reauthentication policies are weak.
Gartner’s advice to buyers is to be wary of any vendor claiming to deliver complete zero trust. That applies to every security company, Fidelis included. Zero trust needs an owner, a defined scope, and metrics of its own, and each tool should be judged by what it adds to the whole.
Gap 2: Verification that happens once
What does your access policy do at 3:00 p.m. to a session it approved at 9:00 a.m.?
In many programs, nothing. The laptop could have been compromised over lunch, or the same account could appear in two countries within the hour, and access would carry on unchanged.
To implement conditional access for zero-trust architecture properly, it helps to separate two things that often get blurred. Continuous access evaluation is the policy side: the access system reassesses an active session instead of relying only on the decision made at sign-in. Continuous threat monitoring is the detection side: security tools watch for suspicious behavior. Monitoring on its own revokes nothing. It only changes access when its findings reach a system that can enforce a decision.
In practice, the link works like this. A risk signal, such as an impossible-travel login, a device falling out of compliance, or unusual access to sensitive data, gets passed to the access policy engine. The engine then reassesses the session against its rules. Acting on that decision requires enforcement capabilities that many environments lack:
Session or token revocation that takes effect before the token naturally expires
Step-up authentication that can be triggered in the middle of a session
Policy controls that can narrow access to specific resources instead of cutting it entirely
Network or endpoint containment for cases where the device itself is the problem
Not every application supports mid-session reassessment, and some only re-check access when a token refresh. Knowing where those limits sit is a part of closing this gap.
Gap 3: Standing privilege
Permanent admin rights may be the oldest gap on this list and still one of the most common. Service accounts keep broad privileges because some automated jobs depend on them. Break-glass accounts sit outside every policy and rarely get reviewed. One compromised privileged account then reaches everything it’s entitled to, at any hour, without anyone approving it.
A just-in-time access zero trust security implementation changes the default. Privilege is requested, approved, time-limited, and logged, and it expires without anyone having to remember. Be careful with approval workflows that approve every request automatically, since that’s standing privilege with an extra click.
Here’s a test you can run this week. Pick one domain admin account and ask what it could reach at 2:00 a.m. on a Sunday with no change ticket open. If the answer matches Monday at noon, the gap is still there.
Gap 4: Rules that pile up, and the systems they skip
Policy debt builds quietly. Firewall rules live in one console, cloud security groups in another, identity policies in a third. Each made sense when someone created it. Years later, nobody can say with confidence which paths from a user to a database are actually allowed.
Three things usually make this worse:
IP-based rules that outlive the project they were written for
Legacy systems exempted “for now” and then forgotten
Workarounds created by people who found the official route too slow
That last one deserves more attention than it gets. When a control makes someone’s job harder, they find another way in, maybe a shared account or an old VPN profile, and that path sits outside your policy entirely. Owners and expiry dates for every exception help. So does treating every workaround as a design problem to fix, rather than a user to blame.
Gap 5: Visibility that ends at the access decision
Zero trust asks you to assume breach, yet many programs fund prevention and leave detection to whatever tools were already there.
A policy engine answers one question: should this request be allowed? It can’t tell you what the session did afterward.
Consider a scenario. A finance analyst’s credentials are stolen through a phishing page. The attacker signs in from the analyst’s own enrolled laptop, which was compromised by the same campaign. MFA is approved, the device passes its compliance check, and the access policy allows the session. Everything the identity layer checks looks right.
Over the next few days, the session starts doing things the analyst never does:
Querying Active Directory for lists of admin accounts and servers
Opening connections to file servers outside the finance segment
Running legitimate admin tools in ways that don’t match the analyst’s usual work
Compressing files and sending them, in small batches, to an unfamiliar external destination over encrypted traffic
No access rule was broken, because the account was entitled to most of this. That’s the line worth drawing:
Identity controls can establishNetwork monitoring and investigation can reveal
Who signed in, and whether MFA passedWhich systems the session actually talked toWhether the device met policy at that momentWhether that traffic fits the user’s normal patternWhether the request was allowedSigns of reconnaissance, lateral movement, and data stagingWhere data went, and how much
East-west traffic inside an approved segment often goes uninspected, and devices that can’t run an agent, like OT equipment, building systems, and older servers, may not show up in any inventory. That’s where lateral movement happens.
This is the gap Fidelis Network® (Fidelis NDR) with Deep Session Inspection® is built for. It inspects traffic across all ports and protocols, including encrypted sessions and files nested inside other files, and maps attacker behavior such as credential theft and lateral movement to the MITRE ATT&CK framework. In the scenario above, that’s the difference between a session that looks legitimate and one that shows reconnaissance, an unusual path between servers, and data leaving the network.
Gartner recommends measuring zero trust by outcomes such as reduced lateral movement of malware. If nobody can see east-west traffic, that outcome can’t be measured either.
Block attacks before damage occurs
Prevent lateral movement inside your network
Reduce false positives & alert fatigue
Gap 6: Cloud workloads that drift
A workload passes every posture check on deployment day. Six weeks later, someone widens a security group to debug a connection and never narrows it again.
Cloud environments change faster than review cycles, so a quarterly assessment may describe infrastructure that no longer exists. Continuous checks against a trusted baseline catch drift in hours instead of at the next audit.
What These Gaps End Up Costing
A single dollar figure would be guesswork. What the data does show is where the risk sits. If a typical strategy mitigates a quarter or less of enterprise risk, most organizations are paying for zero trust while most of their exposure stays outside it.
Gartner also predicted in 2023 that through 2026[4], more than half of cyberattacks would target areas zero trust controls don’t cover. Nobody outside can confirm whether that exact share holds, but the logic is hard to argue with. Attackers go where defenses aren’t.
In practice, the cost shows up as:
A bigger blast radius because one stolen credential reaches further when privilege is permanent.
Longer dwell time, since an intruder with valid credentials looks like an employee.
Spending without matching coverage, with 62% of organizations expecting zero trust to raise costs.
Misplaced confidence, where leaders make decisions based on protection that exists only on a dashboard.
The fourth is the hardest to measure, and the least likely to show up in a budget review.
How Do You Implement Zero Trust in a Network That Already Exists?
Very few organizations get to start from scratch, so the real question is how to implement zero trust security in a network that’s already running.
The most detailed public zero-trust implementation guidelines today come from the NSA, published in January 2026. They begin with a Discovery phase, then Phase One (36 activities) and Phase Two (41 activities). They were written for US defense, but the phases are modular.
For a commercial enterprise, a practical zero trust implementation roadmap looks like this:
Discover users, devices, applications, data, and the flows between them
Prioritize the assets whose compromise would hurt most
Fix identity with phishing-resistant MFA and an owner for every service account
Enforce continuous conditional access, just-in-time privilege, and micro-segmentation
Monitor what happens inside approved sessions
Automate and widen, adding zero trust automation only once policies are proven
Steps one and five get skipped most often, and both come down to visibility.
As for the cost to implement zero-trust, scope drives almost all of it. Affordable approaches to implementing zero trust usually mean configuring features you already own and borrowing from free guidance. NIST SP 1800-35, for instance, documents 19 example zero trust architecture implementation builds using off-the-shelf products.
A zero-trust implementation checklist worth revisiting each quarter:
Inventory covers unmanaged, cloud, and OT assets
Access is re-evaluated during sessions
No permanent admin rights remain
Every exception has an owner and an expiry date
East-west traffic is inspected
Each policy has a metric showing it works
Where Fidelis Security Fits
Fidelis Security doesn’t replace your identity provider, and it isn’t a zero-trust program by itself. Its role is the part many programs underfund: seeing and responding to what happens after access is granted.
Fidelis Elevate® (Fidelis XDR) is where network, endpoint, cloud, and deception signals come together. It maps the cyber terrain across on-premises and cloud environments with risk profiling, discovers and classifies assets, and keeps reassessing endpoints after approval so teams can cut access when behavior changes.
Looking at one piece of the picture doesn’t necessarily provide the full context of what a potential threat is for an organization.
GapWhat Fidelis doesDelivered by
Siloed controlsCorrelates network, endpoint, and other supported security signals to improve investigation context. It doesn’t replace centralized Zero Trust policy management.Fidelis Elevate®One-time verificationMonitor endpoint behavior and suspicious AD activity after authentication. Their signals complement continuous access evaluation.Active Directory InterceptStanding privilegeHelp detect privilege abuse, credential misuse, and suspicious reconnaissance. They don’t replace PAM or just-in-time access controls.Fidelis Deception®Legacy and agentless systemsProvides visibility into traffic and communication paths, including monitored assets without endpoint agents. This supports identifying and investigating policy gaps.Fidelis Network® (Fidelis NDR) with Deep Session Inspection®Missing visibilitySupport network monitoring, behavioral detection, cross-domain correlation, and investigation.Fidelis Network, Endpoint, and ElevateCloud driftHelp identify cloud misconfigurations, unauthorized changes, and configuration drift.Fidelis Halo® (Fidelis CNAPP)
On identity, Skelly adds: “With Fidelis, we can correlate Active Directory activity with network behavior through our NDR product to identify compromised users, privilege escalation attempts, and understand what that user is doing across the network.“
Gaurav Bahadur of Fidelis explains the cloud side: “Continuous drift detection means constantly checking workloads against a trusted baseline. That continuous validation is essential for preventing workloads from becoming vulnerable over time.“
No product gives you absolute resilience, and no zero-trust implementation does either. By Gartner’s numbers, most programs cover about half of what they’re meant to protect. Do you know which half yours covers?
Frequently Asked Questions
How long does a zero-trust implementation take?
Usually several years. NIST’s own reference builds took four, so plan in phases with quarterly milestones.
Who should own the zero-trust program?
Gartner found 59% of initiatives are sponsored by the CIO, CEO, or board. Day-to-day ownership usually sits with security architecture.
Does zero trust replace VPNs and firewalls?
Not right away. Most enterprises retire VPN access gradually, and firewalls stay on as enforcement points.
Is zero trust mandatory for private companies?
Generally no. Most adopt it by choice, or because customers and insurers ask for it.
Which gap should we fix first?
Whichever one protects your most critical asset. For many enterprises, that’s standing privilege on directory and admin accounts.
Citations:
The post Zero Trust Implementation Gaps That Leave Enterprises Exposed appeared first on Fidelis Security.
No Responses