Skip to main content

Governance Gate

The Governance Gate is the fourth stage of the Operating Model Lifecycle and the formal checkpoint where AI-assisted development outputs are reviewed against organizational standards, security requirements, compliance mandates, and quality thresholds before deployment approval. No AI-assisted code SHALL reach production without passing the Governance Gate. This stage operationalizes the governance framework defined in Phase 2: Governance Implementation and enforced through the CI/CD Pipeline.

MANDATORY REQUIREMENT

No AI-assisted code is permitted in production without passing the Governance Gate. This requirement applies to all code, regardless of risk tier, team, or urgency. Emergency hotfixes follow an expedited governance path (defined below) but do NOT bypass governance entirely. Violations of this requirement are treated as Critical policy violations per the Organization-Wide Policy.

Review Criteria

The Governance Gate evaluates AI-assisted work against five mandatory criteria categories. ALL categories MUST pass for the gate to approve deployment.

1. Code Quality Criteria

CriterionRequirementEvidence
All automated tests pass100% pass rateCI/CD test results
Code coverage meets threshold>= organizational standard (typically 80%)Coverage report
Static analysis findings resolvedZero Critical or High findingsSAST scan report
Code review completedAt least 1 approving review from AI-trained reviewerPR approval record
Refactoring completeAll Human Hardening checklist items verifiedHardening checklist in PR

2. Security Criteria

CriterionRequirementEvidence
SAST scan cleanZero Critical or High findingsSAST report
SCA scan cleanZero Critical findings; High findings reviewed and acceptedSCA report
Secrets scan cleanZero findingsSecrets scan report
Security review complete (Tier 3+)Security reviewer sign-offSecurity review record
Data classification complianceNo Restricted data in prompts or codeDLP logs and developer attestation

3. Compliance Criteria

CriterionRequirementEvidence
AI attribution metadata presentAll AI-assisted code has attribution metadataAutomated metadata check
Licensing complianceNo restricted licenses in new dependenciesLicense scan report
Regulatory requirements metApplicable regulatory checks passedCompliance checklist
Policy complianceNo active policy violations for the teamGovernance compliance dashboard

4. Documentation Criteria

CriterionRequirementEvidence
API documentation completeAll new/modified public APIs documentedDocumentation review
Change documentationChange description with scope and impactPR description and/or change ticket
Viability assessment on fileAI Exploration viability assessment documentedViability assessment artifact
Deployment plan documentedDeployment steps, rollback plan, monitoring setupDeployment checklist

5. Risk Criteria

CriterionRequirementEvidence
Risk tier confirmedCurrent risk tier assignment is accurateRisk assessment record
Risk-appropriate governance appliedGovernance level matches risk tierGovernance records
New risks identified and registeredAny risks discovered during hardening are added to the risk registerRisk register entry
Residual risk acceptedRemaining risks are within organizational risk appetiteRisk acceptance record (Tier 3+)

Approval Process

The approval process varies by risk tier to balance rigor with speed:

Tier 1 (Low Risk) — Automated Approval

  • All criteria are evaluated automatically by the CI/CD pipeline
  • If all automated checks pass and a qualified reviewer has approved the PR, the gate passes automatically
  • No manual governance review is required
  • Audit trail is maintained automatically

Tier 2 (Moderate Risk) — Automated + Tech Lead

  • All automated checks must pass
  • Tech Lead explicitly confirms that hardening is complete
  • No additional governance review required unless flagged by automated checks
  • Audit trail is maintained automatically

Tier 3 (Elevated Risk) — Automated + Security + Governance

  • All automated checks must pass
  • Security reviewer has reviewed and approved
  • Governance Lead reviews the complete evidence package and provides approval
  • Approval is recorded in the governance audit trail
  • Typical turnaround: 1-2 business days

Tier 4 (High Risk) — Full Governance Review

  • All automated checks must pass
  • Security reviewer has reviewed and approved
  • Governance Lead reviews the complete evidence package
  • CISO provides final approval
  • Risk acceptance is formally documented
  • Typical turnaround: 2-5 business days
MANDATORY REQUIREMENT

Tier 4 governance reviews MUST NOT be expedited without CISO authorization. The additional review time for Tier 4 work is a deliberate control. Pressure to skip or rush Tier 4 review MUST be escalated to the Steering Committee.

Required Evidence Package

For Tier 3 and Tier 4 reviews, the submitting team MUST provide a governance evidence package containing:

  1. Business Intent Document — The original business intent with success criteria
  2. Viability Assessment — AI exploration viability assessment results
  3. Hardening Checklist — Completed human hardening checklist with all items verified
  4. CI/CD Pipeline Results — Full pipeline execution results showing all gates passed
  5. Security Scan Reports — SAST, SCA, and secrets scan results
  6. Code Review Record — PR review history showing approving review from qualified reviewer
  7. Deployment Plan — Deployment steps, rollback procedure, and monitoring plan
  8. Risk Assessment — Current risk tier with any new risks identified during development

Rejection Handling

When the Governance Gate rejects a submission, the following process applies:

Rejection Process

  1. Rejection notification — The submitting team receives a detailed rejection notice specifying which criteria failed and why
  2. Remediation guidance — The rejecting reviewer provides specific, actionable guidance for addressing each failure
  3. Remediation timeline — A timeline for remediation is agreed upon (typically 2-5 business days for standard issues)
  4. Remediation work — The team addresses the identified issues, returning to Human Hardening as needed
  5. Re-submission — The team re-submits with evidence that all identified issues are resolved
  6. Re-review — The governance reviewer verifies remediation (focused on previously failed criteria)

Common Rejection Reasons

ReasonFrequencyTypical Remediation
Unresolved SAST findingsHighFix identified vulnerabilities; re-scan
Missing AI attribution metadataMediumAdd attribution metadata to all AI-assisted files
Insufficient test coverageMediumWrite additional tests to meet coverage threshold
Missing security review (Tier 3+)MediumRequest and complete security review
Inadequate error handlingMediumAdd specific error handling per hardening checklist
Licensing issues in dependenciesLowReplace problematic dependencies or obtain legal approval

Rejection Metrics

The Governance Lead MUST track rejection metrics:

  • Rejection rate — Percentage of submissions rejected on first attempt (target: < 15%)
  • Rejection reasons — Categorized reasons to identify systemic issues
  • Remediation time — Average time from rejection to successful re-submission
  • Repeat rejections — Submissions rejected more than once (indicates inadequate hardening)

Waiver Process

In exceptional circumstances, specific governance criteria MAY be waived. Waivers are not bypasses — they are documented, time-limited exceptions with compensating controls.

Waiver Requirements

  • Justification — A clear, documented business justification for why the criterion cannot be met
  • Compensating controls — Alternative controls that mitigate the risk of the waived criterion
  • Time limit — Every waiver MUST have an expiration date (maximum 90 days)
  • Approval — Waivers for Tier 1-2 criteria require Governance Lead approval; Tier 3-4 criteria require CISO approval
  • Tracking — All waivers are recorded in the Exception Register and reviewed monthly
MANDATORY REQUIREMENT

Security criteria waivers require CISO approval regardless of risk tier. No waiver MAY be granted for: secrets in code, Restricted data in AI prompts, or deployment without any automated testing. These are absolute requirements with no exception path.

Emergency Governance Path

For genuine production emergencies requiring immediate deployment:

  1. The Engineering Director and Security Lead jointly authorize the emergency deployment
  2. Automated pipeline checks MUST still pass (no bypass of automated gates)
  3. A full governance review MUST be conducted within 24 hours of the emergency deployment
  4. The emergency and its justification MUST be documented in the governance audit trail
  5. The Steering Committee MUST be informed at the next meeting

The Governance Gate is the organization's quality and security conscience. It exists not to slow teams down but to ensure that the speed AI provides does not come at the cost of the standards that protect customers, data, and the organization's reputation. When governance is well-implemented, it becomes a source of confidence, not friction — teams that pass the gate know their code is ready for production.