State-based Policy in Action: Operational Implementation Guide

Phase 1: Compliance & Privacy Guardrails

• Ban PII input: Strictly prohibit staff and students from uploading Personally Identifiable Information (PII)—including student names, histories, report comments, assessment results or attendance logs—into public AI tools.
• Obtain explicit parental consent: Secure verified, opt-in parental consent before allowing students to register for any commercial platform requiring personal details beyond standard school credentials.
• Preserve human accountability: Ensure educators retain final authority; never delegate final student reports, unmonitored grading decisions or official parent communications to automated tools.
• Apply structured risk assessment: Collaborate across faculties to evaluate proposed tools against recognized risk models, such as frameworks from the National AI Centre.

Phase 2: System Infrastructure & Tool Selection

• Prioritise state-owned platforms: Deploy department-sanctioned environments where available—such as NSW’s NSWEduChat or South Australia's EdChat—to ensure student data remains protected within local enterprise boundaries.
• Verify tool safety via ST4S: Check product safety and privacy compliance reports via Education Services Australia’s Safer Technologies 4 Schools (ST4S) portal before procuring third-party EdTech software.
• Block unvetted platforms: Partner with school IT administrators to blacklist unvetted commercial applications, eliminating rogue AI adoption and unauthorized data harvesting.

Phase 3: Curriculum & Pedagogical Design

• Embed ACARA v9 AI literacy: Audit faculty unit plans against ACARA’s Curriculum Connection: Artificial Intelligence to teach computational thinking, algorithms and data analysis explicitly across Mathematics, Technologies and Science.
• Mandate academic integrity training: Require secondary students to complete formal academic integrity modules (such as QCAA's mandatory online certification) to establish clear citation protocols and ethical habits.
• Protect against cognitive offloading: Design tasks that safeguard 'productive struggle', ensuring learners do not use AI generation to bypass foundational domain knowledge, decoding and deep critical thinking.

Phase 4: Assessment Redesign & AI-Resilient Frameworks

• Reject unreliable AI detectors: Cease reliance on automated AI detection tools due to technical unreliability and documented false-positive bias against EAL/D and LBOTE students.
• Increase supervised assessment: Re-establish supervised, in-class checkpoints and handwritten components for high-stakes summative evaluations in line with state authority directives.
• Build in-task checkpoints: Structure assessments around verified process milestones, including sequential draft portfolios, literature mapping, source triage and face-to-face oral defences.
• Incorporate AI critique tasks: Engage students in active evaluation tasks, such as sourcing triage or counter-syntheses, where learners interrogate AI outputs for hallucinations, omissions and algorithmic bias.

Key Support Channels for School Leaders