AI-SDLC Institute™

Source: AISDLC/AI-SDLC-SOPs@3692389 — sops/SOP-1300-01-AI_AI-IRB-Governance-and-Oversight.md

SOP-1300-01-AI_AI-IRB-Governance-and-Oversight
Title: AI-IRB Governance & Oversight

View diagram at source (pin)

Effective Date: (Date of Implementation)
Supersedes: None (New SOP)
Reason for Update: Initial Release
Owner/Document Control: AI Governance Office

Objective

The objective of this Standard Operating Procedure (SOP) is to define and document the governance structure and oversight responsibilities of the AI Institutional Review Board (AI-IRB) within the AI-SDLC framework. This SOP outlines the processes by which the AI-IRB evaluates, approves, and monitors AI projects to ensure compliance with ethical, regulatory, and organizational requirements.

Scope

This SOP applies to all AI initiatives under the jurisdiction of the AI-IRB, including internal product development, external client engagements, research collaborations, and any other AI/ML-based or data-driven endeavors. It also governs the oversight and ethical review of AI models, tools, or algorithms developed or integrated into the production environment by the Performing Organization (Engineering), Contracting Organization (Product/Technical Support), or external partners.

Applicability

All personnel and functional areas—Development, Quality Assurance, Operations, Product, Technical Support, AI Governance Office, and Senior Management—are required to adhere to this SOP when proposing or modifying AI-based projects that fall within the AI-IRB’s purview.

References

Roles and Responsibilities

RoleResponsibility
AI-IRB Chair- Convenes AI-IRB meetings <br/> - Coordinates reviews, ensures conformance to policies <br/> - Approves final decisions
AI-IRB Members- Expert panelists across technical, legal, ethical, operational domains <br/> - Review project proposals, voice concerns, provide recommendations
Project Sponsor- Submits project proposal to AI-IRB <br/> - Addresses feedback/revisions required <br/> - Ensures resources for compliance
AI Governance Office- Maintains repository of AI-IRB decisions <br/> - Tracks open items and follow-ups <br/> - Communicates policy updates
Engineering Dept.- Implements required modifications from AI-IRB <br/> - Submits technical artifacts, ensures compliance with recommendations
Product/Tech Support- Provides user and client perspective <br/> - Coordinates training and user documentation with AI-IRB guidelines
Operations- Implements changes into production environments <br/> - Maintains logs/evidence of compliance post-release
Senior Management- Confirms alignment of AI-IRB with strategic direction <br/> - Resolves escalated issues or cross-organizational conflicts

Procedure Activities

  1. AI-IRB Setup and Membership
    • The AI Governance Office appoints the AI-IRB Chair in consultation with Senior Management.
    • The Chair nominates AI-IRB Members with diverse expertise (ethical, technical, legal).
    • AI-IRB membership is reviewed annually to maintain independence and diversity of opinions.
  2. Project Proposal Submission
    • The Project Sponsor completes the AI-IRB Project Proposal Form describing project scope, objectives, data usage, ethical considerations, and potential impact.
    • The Project Sponsor submits the form to the AI-IRB Chair at least two weeks before the next scheduled meeting.
    • The AI-IRB Secretary (delegated by Chair or AI Governance Office) logs the proposal, assigns a reference number, and distributes it to AI-IRB Members.
  3. Preliminary Screening
    • The AI-IRB Chair designates one or more primary reviewers based on subject matter expertise.
    • Primary reviewers perform an initial analysis:
      • Data privacy compliance (GDPR, CCPA, etc.)
      • Risk classification (low, medium, high)
      • Potential biases (training data, model approach)
      • Ethical risk & impact assessment readiness (SOP-1305-01-AI if applicable)
    • If the preliminary analysis reveals major gaps, the AI-IRB Chair requests clarifications from the Project Sponsor.
  4. AI-IRB Meeting and Discussion
    • The AI-IRB convenes periodically or on-demand (for high-priority projects).
    • The Project Sponsor presents the project proposal with relevant stakeholders (Engineering, Product, etc.).
    • AI-IRB Members discuss potential concerns, ask clarifying questions, and highlight risk areas (e.g., bias, transparency, or safety).
  5. Decision and Post-Meeting Actions
    • The AI-IRB Chair calls for a decision based on the panel’s discussion. Potential outcomes:
      • Approval (with or without minor conditions)
      • Deferred (requires additional info or rework)
      • Rejection (project not viable within current constraints)
    • The AI Governance Office documents the decision and any conditions.
    • The Project Sponsor receives official AI-IRB feedback within two business days.
  6. Tracking and Follow-up
    • The AI-IRB’s conditions (if any) must be addressed before the next milestone.
    • The Engineering Dept. implements changes or mitigations.
    • The Project Sponsor re-submits an updated proposal or relevant artifacts for final sign-off.
    • The AI Governance Office updates the master AI-IRB Decision Registry, logs open items, and sets review timelines.
  7. Deployment Clearance
    • Once the project is near completion in the AI-SDLC, the AI-IRB rechecks if all conditions are met.
    • If satisfied, the AI-IRB issues a formal letter or email clearing the model/project for deployment.
    • Operations, with QA oversight, then proceeds with final release steps (SOP-1220-01 Deployment).
  8. Post-Implementation Review
    • The AI-IRB and the AI Governance Office schedule a post-implementation check (30–90 days post-launch).
    • Observed metrics (model performance, real-world ethical or bias issues, user feedback) are compiled.
    • Lessons learned feed back into future AI-IRB guidelines and best practices.
  9. Annual Review / Ongoing Oversight
    • High-risk or critical AI systems require periodic re-review.
    • The AI-IRB may request updated documentation or performance metrics to ensure ongoing compliance.
    • Non-compliance or unaddressed concerns can trigger corrective actions or suspension of AI system usage.

Forms / Templates

  1. AI-IRB Project Proposal Form
  2. AI-IRB Decision Registry
  3. Change Request / Mitigation Worksheet

Metrics

Records and Archiving

Exemptions

Tools/Software/Technology Used

Change History

VersionDateDescriptionAuthor/Change Owner
1.0(EffectiveDate)Initial ReleaseAI Governance Office
1.1TBDN/AN/A

Approval Signatures

End of SOP

@startuml

’ Define participants (Roles)
participant “Project Sponsor” as PS
participant “AI-IRB Chair” as CHAIR
participant “AI-IRB Members” as MEMBERS
participant “AI Governance Office” as GOV
participant “Engineering Dept.” as ENG
participant “Product/Tech Support” as PTS
participant “Operations” as OPS
participant “Senior Management” as SM

’ 1. Project proposal submission
PS -> CHAIR: Submits AI-IRB Project Proposal Form
CHAIR -> MEMBERS: Distribute proposal for initial screening

alt Major gaps identified
CHAIR -> PS: Request clarifications/further info
else Sufficient detail
MEMBERS -> MEMBERS: Perform preliminary analysis (data privacy, risk classification, etc.)
end

’ 2. AI-IRB Meeting
CHAIR -> CHAIR: Schedule AI-IRB meeting
CHAIR -> PS: Invite Project Sponsor to present
PS -> MEMBERS: Present project scope, objectives, data usage

’ 3. Deliberation
MEMBERS -> CHAIR: Discuss concerns, recommendations
CHAIR -> GOV: Record meeting minutes/decision notes

’ 4. Decision outcomes
alt Approved
CHAIR -> PS: Notifies approval (with or without conditions)
else Deferred
CHAIR -> PS: Notifies deferral; rework required
else Rejected
CHAIR -> PS: Notifies rejection
end

’ 5. Tracking conditions and follow-up
PS -> ENG: Address AI-IRB conditions / implement changes
ENG -> PS: Provide updated compliance or revised proposal
PS -> CHAIR: Submit revised artifacts if needed

’ 6. Final sign-off
CHAIR -> MEMBERS: Final review of revised submission
MEMBERS -> CHAIR: Confirm readiness
CHAIR -> GOV: Update Decision Registry

’ 7. Deployment clearance
CHAIR -> OPS: Issue formal clearance for deployment (if all conditions met)
OPS -> CHAIR: Confirms release scheduling
CHAIR -> PTS: Coordinate end-user training, if needed

’ 8. Post-implementation
GOV -> CHAIR: Schedule post-implementation check (~30-90 days)
CHAIR -> MEMBERS: Evaluate real-world metrics, performance
alt Non-compliance detected
CHAIR -> SM: Escalate major concerns for executive resolution
else All in compliance
note right
Project remains in good standing under AI-IRB oversight
end note
end

@enduml

Short textual explanation:
This sequence diagram illustrates how the AI-IRB Governance & Oversight process unfolds under SOP-1300-01-AI. The Project Sponsor submits a proposal to the AI-IRB Chair, who distributes it to members for initial screening. Following an AI-IRB meeting, members deliberate and issue a decision (approve, defer, or reject). If approved or deferred, the Engineering and Sponsor teams address required conditions, resubmit for final sign-off, and proceed to deployment clearance. A post-implementation review is scheduled to monitor compliance and address any non-compliance issues.