AI Governance Charter Template

01

Section 1

Purpose & Scope

Fill in the bracketed fields below with your organization's specifics. This charter should be signed before any AI pilot goes live, not drafted retroactively after one.

Fields to Complete

[Organization / Program Name] — the entity this charter governs
[In-Scope Use Cases] — the specific AI applications this charter covers (list from your Use-Case Register)
[Out-of-Scope Uses] — applications explicitly excluded, or requiring a separate charter
[Effective Date] & [Review Date] — this charter must have a scheduled review, not an open-ended one

02

Section 2

Approval Authority

Name the role — not a department, an accountable role — authorized to approve each of the following:

Decision	Approving Role
A new AI use case entering pilot	[Role, e.g. Program Director]
A pilot moving to production	[Role, e.g. Steering Committee]
A material change to a production model	[Role, e.g. Technical Committee]
Retiring or decommissioning a model	[Role, e.g. Program Director]

03

Section 3

Human-in-the-Loop Checkpoints

No AI agent or model holds final decision authority over a public resource, a citizen-facing determination, or committed funds. Complete the reviewer column for your program.

What AI Can Do

Draft a first-pass score, recommendation or summary
Flag anomalies for human review
Accelerate document review and comparative analysis

What Stays Human — Named Reviewer

Approving a permit, benefit, penalty or award: [Role]
Accepting a model output as final: [Role]
Committing funds or signing a contract: [Role]

04

Section 4

Data Handling & Bias Monitoring

Data Handling Policy

[Data classification level] permitted for use with external AI providers
[Retention period] for data shared with any AI system
[Named owner] for confidentiality terms in AI vendor contracts

Bias & Error Monitoring

[Review cadence] for testing outputs against demographic/geographic segments
[Threshold] error rate that triggers automatic suspension of a model
[Named owner] accountable for the monitoring log

05

Section 5

Escalation & Review Cycle

Is there a documented process for a frontline user to flag a suspected model error?
Is there a named escalation point reachable within [response time]?
Is this charter reviewed at a fixed interval, not only after an incident?
Is every in-scope use case re-confirmed against this charter at each review?

Next Step

INA's advisory team can help complete and adapt this charter to your program's regulatory context. See also INA's AI Adoption for Public Sector Infrastructure white paper. Request Advisory →