Join our Newsletter — 33% off our NHI Course
Home› Guides› Audit, Assurance & Compliance Top 10: the assurance…
Top 10 Audit, Assurance & Compliance

Audit, Assurance & Compliance Top 10: the assurance failures that sink audits

PDF for newsletter subscribers · free to read online

← All guides
By NHI Mgmt Group, edited by Lalit Choda Updated 14 November 2026 15 min read
On this page

Most failed security audits are not caused by missing controls but by controls that cannot be proven: the evidence is incomplete, the population is wrong, the scope leaves out a system that matters, or the control works on paper and not in practice. This list covers the ten assurance failures that turn into qualified opinions, exceptions and control deficiencies across SOX IT general controls, SOC 1 and SOC 2 examinations, ISO/IEC 27001 certification audits and internal audit reviews. It focuses on how assurance work goes wrong: design versus operating effectiveness, completeness and accuracy of evidence, scoping, reliance on service providers and remediation that is never retested. It is written for control owners, GRC and compliance leads, and internal auditors. For programme-level governance mistakes see the Top 10 GRC Mistakes, and for access-specific findings see the Top 10 Identity Audit Findings.

The Top 10 at a glance

The ten assurance failures that most often turn a working control into an audit exception, with what the auditor tests and the main reference behind each. Each item below explains the failure, why it matters and what to change first.

#FailureWhat the auditor testsMain reference
1Design tested, operation assumedSamples across the whole period, not one walkthroughPCAOB AS 2201 .42 to .45
2Report data with no completeness and accuracy proofHow a system report or extract was produced and checkedPCAOB AS 1105 .10
3Wrong or incomplete populationPopulation source, filters and reconciliation to the system of recordAS 1105 .10; NIST SP 800-53 CA-2
4Scope that misses in-scope systemsSystem boundary, data flows and supporting infrastructureAS 2201 .36; NIST SP 800-53 PL-2
5Service provider reports accepted unreadReport type, period, exceptions and complementary user controlsPCAOB AS 2601; AS 2201 .B17 to .B27
6Review controls without precisionWhat the reviewer looked at, thresholds and follow-upAS 2201 .68
7Evidence that does not show who, when and whatDated, attributable, retained artefactsAS 1105 .05 to .06; AU-6
8Control descriptions that drift from practiceNarrative and frequency against what actually happensAS 2201 .A3; CM-3
9Automated controls relied on without IT general controlsChange, access and operations controls over the systemAS 2201 .47, .B29 to .B33
10Remediation closed without retestingOperating time of the fixed control and retest evidenceAS 2201 .53; CA-5

The references are PCAOB auditing standards for audits of public companies and NIST SP 800-53 Rev. 5 assessment controls. The same failure modes appear under AICPA attestation standards for SOC reports and in ISO/IEC 27001 clause 9.2 internal audits, so the fixes carry across.

1. Design tested, operation assumed

The issue: the team shows the auditor a well-written procedure and one good example, and treats that as proof the control works. Design effectiveness and operating effectiveness are different questions.

Why it matters: PCAOB AS 2201 separates them. Design effectiveness asks whether the control, operated as prescribed by people with the right authority and competence, would meet its objective (paragraphs .42 to .43). Operating effectiveness asks whether it actually operated that way, performed by a competent and authorised person (.44 to .45). A walkthrough of one transaction can confirm design, but a SOC 2 Type 2 report, a SOX audit or an internal audit of operating effectiveness needs evidence across the period. AS 2201 also ranks procedures from weakest to strongest: inquiry, observation, inspection, re-performance, and states that inquiry alone is not sufficient (.50).

What auditors test: a sample of occurrences spread through the period, sized to the control's frequency and risk, often re-performed.

What to do: decide for every key control what an occurrence is and what proves it happened, then keep that evidence every time, not only before fieldwork. A control with no evidence for three months is a deviation for three months. See the Audit Exception vs Control Deficiency.

2. Report data with no completeness and accuracy proof

The issue: a control depends on a system report, such as a list of terminated users, a change log export or a vulnerability scan summary, and nobody can show that the report contains every relevant record and that its fields are correct. Auditors call this information produced by the entity (IPE).

Why it matters: AS 1105 paragraph .10 requires the auditor, when using company-produced information as evidence, to test its accuracy and completeness, or test the controls over them, and to judge whether it is precise and detailed enough. If the report is wrong, every control that consumes it is unproven, however carefully the reviewer worked. A saved query that silently drops disabled accounts, or a spreadsheet edited by hand before review, breaks the chain.

What auditors test: the query or report parameters, a screenshot or log of the run with date and user, record counts reconciled to the source, and evidence the output was not altered.

What to do: keep an IPE register listing each report, its source system, the query or standard report name, the parameters and the reconciliation step. Prefer standard, unmodified system reports; where custom queries are needed, put them under change control and capture the run evidence automatically. See the Audit Evidence and Continuous Controls Monitoring Guide.

3. Wrong or incomplete population

The issue: the sample is drawn from a population that is not the full set. Typical cases: changes pulled only from the ticketing tool when some deployments bypass it, users pulled from one directory when an application has local accounts, or a date filter that misses the first and last weeks of the period.

Why it matters: a clean sample from an incomplete population proves nothing about the items left out. An auditor who finds production changes in deployment logs with no matching ticket will usually raise a finding on the change control itself, not just on the evidence.

What auditors test: where the population came from, how it was filtered, and whether it reconciles to an independent source (deployment pipeline logs against tickets, HR leavers against directory disables).

What to do: define the population source for each control in writing, reconcile it to a second source before the sample is drawn, and let the auditor select the sample. For change management, compare the deployment system's record of production releases with approved tickets every month. See the SOX IT General Controls (ITGC) Guide.

4. Scope that misses in-scope systems

The issue: the audit boundary covers the main application but leaves out the systems that can change its behaviour or data: the identity provider, the CI/CD pipeline, the database layer, the cloud account, the job scheduler or the service accounts that integrate them.

Why it matters: AS 2201 paragraph .36 requires the auditor to understand how IT affects the flow of transactions, so a boundary drawn too tightly is often challenged during planning and expanded late, with no evidence ready. In SOC 2 work the system description defines what users of the report can rely on: anything outside it is simply not assured, however much the customer assumes otherwise.

What auditors test: data flow diagrams, the inventory of supporting infrastructure, and who can change code, configuration and data in each layer.

What to do: scope from the data flow outwards. List every system that stores, processes or can alter in-scope data, every identity (human and non-human) that can change it, and every provider that hosts it, and record why anything is excluded. NIST SP 800-53 PL-2 asks for the system boundary to be described in the system security plan; use that as the scoping record. See the Audit Readiness for AI Agents and NHIs.

5. Service provider reports accepted unread

The issue: a vendor's SOC report is filed as evidence without checking its type, period, scope, exceptions or the controls it expects the customer to run.

Why it matters: PCAOB AS 2601 explains that a Type 1 report covers whether controls were suitably designed and placed in operation at a date, and gives no opinion on operating effectiveness, so it cannot support reduced control risk. A Type 2 report adds tests, results and exceptions over a period, which should ordinarily be at least six months (.53), and shorter periods or older tests give less support (.16). AS 2601 also notes that the service organisation's controls may assume the user organisation runs certain controls of its own (.31, .46): in SOC terminology these are complementary user entity controls (CUECs). AS 2201 adds that a report without tests and results does not evidence operating effectiveness (.B20 note).

What auditors test: that you read the report, followed up exceptions, covered any gap between the report period and your own period (often with a bridge letter from the provider), checked how subservice organisations were handled, and evidenced your own CUECs.

What to do: review every critical provider's report with a fixed checklist: type, period, system description, carved-out subservice organisations, exceptions and the provider's response, CUECs, and the opinion. Map each CUEC to a named internal control and owner. NIST SP 800-53 SA-9 and SR-6 cover external services and supplier assessments. See the How to Read a SOC 2 Report Guide.

6. Review controls without precision

The issue: management review controls (a monthly review of privileged access, a quarterly firewall rule review, a review of security alerts) are evidenced by a signature or an email saying "reviewed, no issues".

Why it matters: a review is only as strong as what the reviewer actually examined. AS 2201 paragraph .68 says compensating controls count only if they operate at a level of precision sufficient to prevent or detect the issue. The same test applies to the review itself: an auditor cannot re-perform a sign-off that records no criteria, no thresholds and no investigated items.

What auditors test: the review criteria, the data reviewed (back to point 2), items flagged, how they were followed up, and whether the reviewer was independent of the activity.

What to do: write each review with explicit criteria (for example: every new member of an admin role, every rule allowing any source on a management port), record each item flagged and its outcome, and block self-review. A review that finds nothing for two years should itself be questioned.

7. Evidence that does not show who, when and what

The issue: screenshots with no date or system name, approvals given in chat, tickets approved after the change was deployed, and evidence collected weeks after the event or recreated for the audit.

Why it matters: AS 1105 describes appropriate evidence as relevant and reliable (.06), and sufficiency as the quantity needed, which rises with risk and falls as quality improves (.05). Undated or reconstructed artefacts are low quality, so the auditor needs more of them or rejects them. Evidence that the approval came after the deployment is a deviation, not a timing quirk.

What auditors test: timestamps, the identity of the approver or performer, the system of origin, and the sequence of events.

What to do: capture evidence at the moment the control operates and from the system itself (workflow approvals, immutable logs, signed exports), keep it in one repository with retention covering the audit period plus the auditor's needs, and track requests with a register. NIST SP 800-53 AU-6 covers review of audit records. See the Audit Evidence Request Tracker Template.

8. Control descriptions that drift from practice

The issue: the control narrative says "weekly", the team does it monthly; the narrative names a tool that was replaced last year; the policy requires two approvers and the workflow allows one.

Why it matters: the auditor tests against the description. If practice differs, it is a deviation even when the real practice is reasonable. Under AS 2201 Appendix A, a design deficiency is a control that is missing or not properly designed, and an operating deficiency is one that does not operate as designed (.A3): drift produces one or the other.

What auditors test: the narrative, frequency, owner and tool, compared with the evidence produced.

What to do: review control descriptions with owners before every audit period starts, tie tool and process changes to an update of the control record through change management (NIST SP 800-53 CM-3), and write narratives at the level of detail you can always meet. See the Compliance Automation Guide.

9. Automated controls relied on without IT general controls

The issue: the organisation relies on a system-enforced control (a configured approval limit, an automatic account lock, an enforced MFA policy) but cannot show that changes to that configuration, and access to change it, are controlled.

Why it matters: AS 2201 says automated controls generally carry lower risk when the relevant IT general controls are effective (.47). Benchmarking, testing an automated control once and relying on it later, is allowed only when general controls over program changes, access and operations are effective and tested and the control is verified unchanged, and the baseline must be reset periodically (.B29 to .B33). Weak change or access controls remove that reliance and force full testing.

What auditors test: who can change the configuration (including service accounts and pipeline identities), change tickets for the period, and configuration snapshots compared with the baseline.

What to do: list automated key controls, export their configuration on a schedule, alert on drift, and restrict who and what can change them. See the SOC 2 CC8 Change Management Guide.

10. Remediation closed without retesting

The issue: a finding is marked closed when the fix is deployed, not when the fixed control has operated long enough to be tested and has passed.

Why it matters: AS 2201 allows a new control to replace a superseded one only if it has operated long enough to be tested (.53). A fix made two weeks before period end may not count, and a repeat finding the next year is usually rated more severely. Severity depends on the reasonable possibility of failure and its potential magnitude, not on whether a failure actually occurred (.64).

What auditors test: the remediation date, the occurrences since then, and independent retest evidence.

What to do: track every finding in a plan of action and milestones with an owner, root cause, target date and a separate "validated" status set only after retest by someone independent of the fix (NIST SP 800-53 CA-5 and CA-2(1)). See the Plan of Action and Milestones (POA&M) Guide.

Mapping to control frameworks

Where each failure sits in common frameworks. NIST and CSF identifiers were checked against the official NIST datasets.

FailureNIST SP 800-53 Rev. 5NIST CSF 2.0ISO/IEC 27001:2022
1, 3, 7: testing and evidenceCA-2 Control Assessments, AU-6ID.IM-01Clause 9.2 Internal audit
2: report dataCA-2, CA-7 Continuous MonitoringID.IM-01Clause 9.1 Monitoring and measurement
4: scopePL-2 System Security and Privacy PlansGV.OV-03Clause 4.3 Scope of the ISMS
5: service providersSA-9, SR-6GV.SC-07Annex A 5.19 to 5.22 supplier controls
6, 8: review and descriptionCA-7, CM-3GV.PO-02Clause 9.3 Management review
9: automated controlsCM-3 Configuration Change ControlID.IM-02Annex A 8.32 Change management
10: remediationCA-5 Plan of Action and Milestones, CA-2(1)ID.IM-01Clause 10.2 Nonconformity and corrective action

The IIA's Global Internal Audit Standards (2024) cover the internal audit side, including the principle "Communicate Engagement Results and Monitor Action Plans", which applies directly to failure 10. See the Internal Audit for Cybersecurity Guide.

Where to start

  1. List key controls and their evidence: for each, record the occurrence, frequency, owner, population source and the artefact that proves it operated (failures 1, 3, 7).
  2. Build the IPE register: every report a control depends on, with its parameters and reconciliation step (failure 2).
  3. Redraw the scope from data flows: include identity, pipeline, hosting and the non-human identities that can change in-scope systems (failure 4).
  4. Read every critical provider report: check type, period and exceptions, and map every CUEC to an internal owner (failure 5).
  5. Rewrite review controls with criteria: what is reviewed, what triggers follow-up, and how outcomes are recorded (failure 6).
  6. Align narratives before the period starts: then link control records to change management (failures 8, 9).
  7. Close findings only on retest: independent validation after the control has operated (failure 10).

Measure progress with a small set of indicators: the share of key controls with evidence captured automatically, the number of IPE reports with a documented reconciliation, the share of critical providers with a reviewed current report and mapped CUECs, the repeat finding rate year on year, and the average time from fix to validated closure. For the wider field, see the Audit, Assurance and Compliance Guide.

Standards and references

Related NHI Mgmt Group resources: Audit, Assurance and Compliance Guide · Top 10 GRC Mistakes · Top 10 Identity Audit Findings · SOC 2 Guide · ISMS Internal Audit and Management Review

PDF for subscribers

Take this guide with you

Every section, table and checklist in one PDF, with links to the sources and related guides. Subscribe to our weekly NHI & AI Identity Journal and we’ll email you a download link.

Free weekly newsletter

Subscribe to the NHI & AI Identity Journal

The latest on NHI and Agentic AI security – articles, research, breaches, news and events every week.

Bonus 33% off our NHI Course when you subscribe.

NHIMG Editorial Note
Researched and written by the NHI Mgmt Group editorial team using AI tools, under the editorial direction of Lalit Choda. Facts are drawn from the sources listed on this page and checked by our automated quality process. Last updated 14 November 2026. Spotted an error? Use our contact page.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM and Agentic AI security. nhimg.org

PDF Downloads for Newsletter Subscribers

Get the guide

Enter your details and we’ll email you a download link.

The full guide is free to read online. The PDF is a thank you for subscribing.

Fill in all fields marked * and tick the PDF box to continue.