Navigation
    Add a header to begin generating the table of contents

    AI Operations Procedure

    1. Preamble

    This procedure sets out the mandatory steps staff, students and affiliates must follow when procuring, deploying or using Artificial Intelligence (AI) tools at Cairnmillar Institute.

    2. Procedures

    2.1. What staff and students must NOT put into AI tools

    Unless the tool is approved and listed on the AI Register:

    Do not enter:

    • Client notes or clinical information
    • Student personal information
    • Staff personal information
    • Sensitive organisational data
    • Research data not approved for AI processing
    • Any information covered by privacy or confidentiality obligations.

    De-identified examples may be used for learning or drafting.

    2.2. When a DPIA / PIA is mandatory

    Before using a new AI tool or new AI use case, a DPIA/PIA must be completed if the AI use involves:

    • Personal, student or health information
    • Monitoring or profiling individuals
    • Automated or semi-automated decision support
    • Use with minors
    • New data flows outside existing systems.

    Approval must be obtained from the Technology Governance Committee before use.

    2.3. Recording AI involvement in decisions (legal protection clause)

    Where AI assists in work that influences decisions about students, clients, staff or governance matters, the staff member must keep a local record noting:

    • The AI tool used
    • What task AI assisted with
    • What human review occurred
    • The final decision made by the staff member.

    2.4 Approved vs Non-approved AI tools

    Only AI tools:

    • Approved by the Technology Governance Committee, and
    • Listed on the AI Register
    • may be used for operational, clinical, student, research or administrative work.

    Staff should preferentially use organisation-approved AI tools for work purposes where available.

    In circumstances where no approved tool exists, staff may use personally procured AI tools for low-risk work activities, provided that:

    • no confidential, sensitive, proprietary, personal, student, client, financial, or commercially sensitive information is entered into the system;
    • no intellectual property or unpublished organisational materials are uploaded unless expressly authorised;
    • outputs are reviewed and verified by the staff member prior to use;
    • the use complies with organisational privacy, cybersecurity, records management, and confidentiality obligations; and
    • the tool is not used for high-risk decision-making, assessment, evaluation, or automated actions on behalf of the organisation.

    Staff must not independently procure AI tools that require integration with organisational systems, access to organisational data repositories, or use of organisational credentials without prior approval from the relevant delegate.

    2.5. Research Use of AI

    Researchers using AI tools must:

    • comply with the Research Ethics Policy, Research Integrity and Conduct Policy, and applicable HREC requirements;
    • disclose AI use in ethics applications, amendments, publications, and theses where appropriate;
    • ensure participant consent and data handling arrangements permit AI-assisted processing;
    • not upload identifiable or sensitive research data into non-approved AI tools;
    • verify the accuracy and integrity of AI-assisted outputs;
    • maintain records of significant AI involvement in data analysis, coding, drafting, or interpretation; and
    • ensure AI tools are not used in ways that compromise reproducibility, authorship integrity, or research transparency.

    Introduction of AI tools into approved human research projects may require HREC review or amendment approval prior to use.

    3. Responsibilities

    3.1. Staff and Students

    • Complete AI training
    • Follow this procedure
    • Use AI ethically and safely.

    3.2. Managers

    • Ensure staff understand and follow this procedure
    • Escalate new AI uses for approval

    3.3. Technology Governance Committee

    • Maintain the AI Register
    • Review Data Protection /Privacy Impact Assessments (DPIAs)
    • Approve AI tools
    • Report to Executive and Council

    4. Monitoring

    Cairnmillar may monitor usage of approved AI tools to ensure compliance with privacy, security and policy obligations.

     

    5. Breach

    Failure to follow this procedure may result in disciplinary action and, where relevant, privacy or clinical reporting obligation.