ADR 011: AI Tool Governance

Status: Accepted | Date: 2025-08-15

Context

Generative AI tools used for development and operations can process sensitive data and make automated decisions affecting security, privacy, and compliance. Without governance ensuring human oversight, these tools pose significant risks including unauthorized data exposure, biased decision-making, and compliance violations.

High-risk scenarios include:

• Automated Decision-Making: AI tools making policy, approval, or resource allocation decisions without human review
• Government Data Processing: Sensitive organisational data processed by offshore AI services
• Uncontrolled AI Outputs: AI-generated content, code, or analysis used without human validation
• Privacy Violations: Personal information processed by AI without appropriate consent or controls

References:

• ACSC Information Security Manual (ISM)
• WA Cyber Security Policy
• Privacy Act 1988
• WA Government Artificial Intelligence Policy and Assurance Framework
• Linux Foundation Agentic AI Foundation
• Oxide RFD 576: Using LLMs at Oxide - values-based approach to AI tool governance

Decision

Implement mandatory human oversight for all AI tool usage with pre-approval for any AI tools that process organisational data or generate outputs used in official capacity.

Human Oversight Requirements:

Adopt a values-based approach to AI governance (per Oxide RFD 576):

• Responsibility: Humans bear responsibility for AI-generated artifacts - the tool acts at human behest
• Rigor: AI should promote and reinforce rigorous thinking, not replace it with generated content
• Output Validation: All AI-generated content must be reviewed by qualified humans before use
• Decision Accountability: Clear human responsibility for all AI-assisted decisions

Covered AI Tools:

This ADR applies to all AI tools including:

• Development and coding assistance tools
• Content generation and writing assistants
• Data analysis and business intelligence platforms
• Automated testing and code review tools

Requirements:

AI tools must not:

• Automatically perform actions in customer-facing environments or on production infrastructure
• Process sensitive data with third parties without a formal contractual arrangement
• Automatically merge or release changes without human review

AI tools must:

• Run in isolated or local environments (refer to ADR 001: Application Isolation) with minimal permissions
• Have no direct network access to internal systems or databases
• Include technical guardrails against data exfiltration

Implementation Examples:

• Endorsed: Developer tooling with human code review for all generated code before merge
• Rejected: Automated tools that merge pull requests or deploy without human approval

Strategic Research

The following AI-assisted security tools are under investigation for potential future adoption:

These tools are being evaluated for alignment with the human oversight requirements outlined in this ADR. Any adoption will require demonstrated compliance with mandatory requirements above.

Consequences

Benefits:

• Ensures human accountability for all AI-assisted decisions
• Maintains compliance with Privacy Act and data sovereignty requirements
• Prevents automated actions in production environments without approval
• Establishes clear audit trail for responsible AI usage

Risks if not implemented:

• Unauthorized data exposure to offshore AI services
• AI making critical decisions without human oversight
• Compliance violations and regulatory breaches
• Operational errors from unchecked AI outputs