GAO's finding: Federal zero-trust plans remain incomplete
The U.S. Government Accountability Office (GAO) identified a significant gap between federal agencies' adoption of zero-trust cybersecurity planning and the completeness of those plans.
In its November 18, 2026 report, Cybersecurity: Agencies Need to Fully Address Federal Guidance as They Refine Zero Trust Architecture (GAO-27-108661), GAO examined the zero-trust implementation plans of 22 civilian agencies covered by the Chief Financial Officers Act.
As of August 2026, only five agencies had addressed all the selected Office of Management and Budget (OMB) and National Institute of Standards and Technology (NIST) guidance examined by GAO. The remaining 17 had not fully addressed that guidance.
GAO identified planning deficiencies involving the identification of actors and assets, documentation of implementation responsibilities and processes, expansion of zero-trust architecture, and assessment of security maturity for high-value assets and high-impact systems.
GAO issued 40 recommendations to 17 agencies. Thirteen agencies agreed with the recommendations; four neither agreed nor disagreed.
The report does not establish that those agencies had suffered a cyberattack because of the identified planning gaps. Its warning is prospective: incomplete planning can impair agencies' ability to allocate resources, establish and maintain zero-trust architecture, and protect federal systems and operations.
That finding raises a related question for organizational information governance: How can an agency reliably enforce access decisions when its knowledge of the assets, identities, authorities, and relationships involved is incomplete or outdated?
GovKM's perceived Organizational Information Continuity challenge
Zero trust assumes that access should be evaluated using reliable information about the requesting actor, the target resource, and the conditions governing access.
GovKM's analysis is that the quality of those decisions depends partly on Organizational Information Continuity: the ability to preserve, maintain, and reconstruct authoritative information and its relationships as systems, personnel, policies, and organizational responsibilities change.
GAO evaluated zero-trust planning against selected federal guidance. It did not conduct a GovKM continuity assessment or establish the following inferred failures as actual incidents. They are operational risks that agencies should evaluate while addressing GAO's recommendations.
Challenge 1: An asset inventory may not preserve the information needed to govern the asset
Identifying an application, repository, dataset, or device is a necessary starting point. It does not necessarily establish the asset's business owner, information sensitivity, authoritative purpose, dependencies, permitted uses, or applicable retention and access requirements.
When those relationships are maintained in separate systems or depend on individual employees' knowledge, a security team may struggle to determine whether an access rule still reflects the asset's current operational role.
The consequences extend beyond cybersecurity. Migration, consolidation, reassignment, and retirement decisions may be made without a complete account of what the asset supports or which information obligations survive the change.
Consequence for AI trustworthiness: An AI assistant connected to enterprise information may retrieve content from an asset without sufficient evidence that the content is authoritative, current, or appropriate for the requesting user. An access-control system may correctly enforce a configured permission while the underlying classification or entitlement is outdated. The resulting AI answer could be plausible but improperly scoped or based on superseded information.
This is a prospective risk, not a claim that GAO found such an AI failure.
Challenge 2: Identity verification does not, by itself, establish current decision authority
Zero trust requires attention to actors, including people, devices, services, and workloads. But confirming an actor's identity is different from establishing the authority that actor currently holds.
An employee may change roles. A contractor's authorization may expire. A service account may retain permissions after the application it supported changes. A delegated approval may be valid only for a defined period or purpose.
If identity records, organizational assignments, delegations, and access decisions are not kept consistent, the agency may be unable to reconstruct why an entitlement was granted or whether it should remain effective.
Consequence for AI trustworthiness: An AI system acting on behalf of a user or service could receive information under an entitlement that no longer reflects current authority. Conversely, legitimate work could be blocked because the system cannot establish an updated delegation. In either case, authentication alone does not demonstrate that the AI's access or resulting action was properly authorized.
Challenge 3: A zero-trust plan may document intended controls without preserving evidence of their implementation
GAO's recommendations address not only technical migration steps but also implementation responsibilities, processes, and maturity assessments.
A plan describes what an agency intends to accomplish. Operational assurance requires evidence of what was configured, tested, approved, changed, and reviewed.
If implementation records are fragmented, a subsequent security officer may see a completed milestone without being able to determine which systems were covered, which exceptions remained, who accepted residual risk, or whether later changes invalidated the earlier assessment.
Consequence for AI trustworthiness: An AI assistant used to summarize security posture could mistake an approved plan for completed implementation, treat an old maturity assessment as current, or omit an unresolved exception. Without evidence-linked status and effective dates, the agency may be unable to verify the assistant's conclusion.
Again, these are GovKM's inferred failure pathways, not AI incidents documented by GAO.
GovKM's proposed solution: Govern the information that zero trust depends on
GovKM recommends treating zero-trust implementation as both a cybersecurity program and an information-governance problem.
The objective is not to replace existing identity, security, or asset-management platforms. It is to make the information supporting their decisions authoritative, connected, current, and reconstructable.
1. Establish a governed inventory of actors, assets, and authoritative sources
Begin with the agency's existing identity directories, asset inventories, configuration records, information repositories, security classifications, and system authorization records.
For each relevant asset, establish its accountable owner, authoritative source record, operational purpose, sensitivity, dependencies, applicable policies, and lifecycle status. For each actor, establish the authoritative identity source and the process by which roles and entitlements are granted, reviewed, changed, and revoked.
NIST SP 800-207, Zero Trust Architecture, describes zero-trust principles and migration considerations, including the need to understand enterprise assets and subjects before applying policy decisions.
For federal agencies, OMB Memorandum M-22-09, Moving the U.S. Government Toward Zero Trust Cybersecurity Principles, established federal zero-trust strategy requirements with an end-of-fiscal-year 2024 implementation milestone. Agencies must evaluate its applicability alongside subsequent directives, agency-specific requirements, and the current implementation baseline.
The inventory should be governed as an operational source of evidence, not simply a spreadsheet assembled for an assessment.
2. Connect identities and assets to their governing authority
An access decision should be explainable through the current policy, approved role or delegation, asset classification, applicable conditions, and accountable decision authority.
The agency should document how these relationships change when an employee transfers, a contractor departs, an application is replaced, or a policy is superseded.
44 U.S.C. § 3554 establishes federal agency responsibilities for information security programs, including risk-based protection of agency information and systems.
NIST SP 800-53 Rev. 5 provides a catalog of security and privacy controls relevant to access control, account management, identification, authentication, configuration management, and auditability. Control selection and implementation must follow the agency's applicable security categorization, authorization process, and governing requirements; the catalog is not a declaration that every control applies identically to every system.
GovKM's proposed continuity method would preserve the relationship between an entitlement, its approval, its effective period, and the evidence supporting it.
3. Preserve implementation decisions and their evidence
For each zero-trust milestone, retain a traceable record of the responsible official, affected systems, planned control, implemented configuration, verification result, exceptions, corrective actions, and approval decision.
Where an implementation decision changes, the prior state should remain reconstructable without being mistaken for the current operative state.
OMB Circular A-130, Managing Information as a Strategic Resource, provides federal policy direction for managing information resources and information security responsibilities.
Federal records obligations also matter. 44 U.S.C. § 3101 requires agency heads to make and preserve records adequately documenting agency organization, functions, policies, decisions, procedures, and essential transactions. 36 CFR § 1222.22 addresses the records agencies must create and maintain to document their activities.
These authorities support the need for adequate documentation. They do not mean that every security event or technical log must be retained permanently. Applicable schedules, security requirements, and records determinations govern retention and disposition.
4. Make changes to authority and security posture traceable over time
Zero trust is not a one-time configuration exercise. Systems change, identities change, and access policies must be reevaluated.
GovKM recommends recording the effective dates and supersession relationships for policies, entitlements, system ownership, security assessments, and approved exceptions.
A reviewer should be able to establish not only the current rule but also which rule was operative when a particular access decision occurred.
The CISA Zero Trust Maturity Model, Version 2.0, provides a framework for assessing maturity across identity, devices, networks, applications and workloads, and data, with visibility and analytics, automation and orchestration, and governance as cross-cutting capabilities.
The maturity model can inform assessment and improvement. Its use must be distinguished from a legal requirement, and an agency should identify which specific guidance or direction governs its own implementation.
5. Prepare the environment for governed AI access
AI introduces another class of actors and information consumers into an already complex environment.
Before connecting an AI assistant or agent to agency repositories, establish which identity it operates under, what data it may retrieve, what actions it may perform, and which records establish the authority for those permissions.
Responses that materially inform decisions should be capable of being checked against accessible, authoritative source records, including relevant versions and effective dates.
GovKM's proposed OrgAI architecture is intended to support continuity among sources, authority, evidence, decisions, actions, and records as they change. This is a proposed application of GovKM's continuity approach, not a claim that OrgAI has been independently validated or deployed as a federal zero-trust enforcement system.
Existing agency identity providers, policy enforcement mechanisms, cybersecurity platforms, records systems, and human decision authorities remain responsible for their respective functions.
How an agency can verify that the solution works
An implementation should be evaluated through observable tests rather than the existence of completed plans alone.
An agency could select representative high-value assets and high-impact systems, then attempt to reconstruct:
- The authoritative inventory record and accountable owner for each selected asset.
- The identities, roles, delegations, and policies governing access.
- The evidence supporting the latest entitlement and classification decisions.
- The current control configuration and its most recent verification.
- The history of exceptions, corrective actions, and superseded decisions.
- The records that would allow a new responsible official to understand and continue the work.
For AI-enabled retrieval, testing should additionally determine whether the system respects the requesting actor's current permissions, distinguishes current from superseded information, identifies the sources supporting material assertions, and discloses uncertainty when authoritative evidence is missing.
A successful result would demonstrate that selected decisions can be independently reconstructed and tested. It would not, by itself, establish enterprise-wide compliance, eliminate cyber risk, or guarantee AI accuracy.
The same continuity problem exists beyond federal cybersecurity
The federal requirements are specific to their legal and administrative context. The underlying information problem is not.
A financial institution replacing an identity-management platform must preserve entitlement ownership, approval history, and regulatory access requirements. A healthcare organization integrating clinical systems must establish which users, applications, and data classifications govern sensitive records. A company deploying enterprise AI must prevent retrieval from becoming a shortcut around existing authorization boundaries.
Each organization operates under its own governing requirements. The reusable method is to connect authoritative information, ownership, access decisions, change history, and verification evidence so those relationships survive organizational and technical transitions.
What GovKM can do
GAO's findings demonstrate why a zero-trust implementation plan must address the federal guidance relevant to the systems being protected.
GovKM's additional conclusion is that sustainable zero trust also depends on the continuity of the information supporting security decisions. An agency should be able to establish what an asset is, who governs it, who is authorized to use it, which policy applies, and what evidence supports the current state—even after systems, people, and policies change.
GovKM can help organizations assess those information relationships, identify continuity gaps, prepare governed source and authority records, and design verification procedures that support cybersecurity implementation and trustworthy AI adoption.
Contact GovKM to discuss an Organizational Information Continuity assessment or implementation engagement.



