Human-led, AI-assisted, Policy-compliant, Privacy-first
USMS's approach to AI in schools is anchored in the Australian Framework for Generative AI in Schools -built compliant by design, never AI-first by default.
AI That Serves Education -On Australian Terms
We treat AI in schools as a policy-led architecture problem -not a technology showcase. Every AI feature in USMS maps directly to the six mandatory principles of the Australian Framework for Generative AI in Schools.
If a feature can't be justified against a principle, it doesn't ship.
Six Principles. Non-negotiable.
Human Oversight
Teachers remain in control. AI suggests, humans decide. Every consequential output is reviewed by a qualified educator.
Transparency
AI-generated content is clearly labelled. Confidence scores and rationale are surfaced so users understand the why behind every recommendation.
Fairness
Continuous bias monitoring across cohorts. Models are tested against equity criteria before deployment and audited regularly.
Accountability
Full audit trails of prompts, outputs and AI-influenced decisions. Clear ownership at the school, district and platform level.
Privacy & Security
Aligned with the Privacy Act 1988. No personal or sensitive student data is used to generate AI analysis. Encrypted at rest and in transit.
Social Wellbeing
AI is deployed only where it improves teaching, learning and wellbeing -never to replace human relationships in education.
How AI Behaves Inside USMS
AI supports teachers -it does not replace them
No sensitive or personal student data is used for AI analysis
Controlled AI outputs only -no free-flow generation
Teacher approval required for every AI-influenced decision
Where AI Genuinely Helps
Teacher Copilots
AI assists with lesson plans, rubrics and resource suggestions - always reviewed and approved by the teacher.
Writing Feedback
Formative writing suggestions on grammar and structure. The teacher remains the final assessor.
Early Warning Signals
Pattern-based attendance and engagement signals to support pastoral care -non-identifying and human-reviewed.
Administrative Automation
Drafting communications, scheduling and routine workflows. Humans remain in the approval loop.
What USMS Will Never Do
Automated grading without human validation
Behavioural or emotional profiling of students
Facial recognition or surveillance
Black-box decision making
Compliance Built Into the Platform
AI Usage Logs
Every prompt, output and AI-assisted decision is logged for review.
Audit Trails
Tamper-evident audit trails meeting Australian compliance requirements.
Role-Based Access
Granular AI permissions per role -student, teacher, admin, principal.
Consent-Based Data Usage
Explicit consent management with clear data retention and right-to-access policies.
A Layered, Auditable AI Architecture
Data Layer
Hosted in Australia with full data sovereignty. Encrypted at rest and in transit. Role-based access controls.
AI Layer
Controlled, fine-tuned models. No direct exposure to public LLMs. All prompts logged and filtered.
Governance Layer
Explainability dashboards, bias monitoring and continuous model performance tracking.
Human-in-the-Loop
Teachers approve AI outputs. No autonomous decision-making in any consequential workflow.
"Human-led, AI-assisted, policy-compliant, and privacy-first."
That's not a tagline. It's the architectural constraint behind every AI capability we ship.