Trust, Security & Scientific Assurance
Australian Child-Safety Infrastructure · Last updated: 27 October 2026
SAFE provides upstream early identification, structured risk assessment, and child-safeguarding governance to help early childhood services and organisations recognise, record and act on child-safety concerns. Reflecting Australian Institute of Family Studies (AIFS) best practice, which pairs empirical risk assessment with professional clinical judgement, our architecture combines onshore Australian cloud hosting, fine-grained access privileges, deterministic decision-support science and Human-in-the-Loop (HITL) Expert Case Review.
Australian Core Hosting
Core child records, SAFE Assessments and Expert Case Review records are hosted onshore on AWS in Sydney, Australia.
Encryption & Fine-Grained Access
Protected in transit using TLS, encrypted at rest, and strictly scoped by organisation, service and fine-grained user privileges.
Deterministic Triage Science
Structured Random Forest risk prioritisation with zero free-text ingestion and zero public AI training.
Human-in-the-Loop (HITL) Review
Every submitted assessment receives human Expert Case Review by experienced child-protection specialists.
Research Lineage, Clinical Stewardship & Support
SAFE's scientific method and child-protection framework were developed in Australia through a multi-year collaboration spanning national research institutions, clinical child-protection practitioners and social-impact funding partners.
CSIRO & Data61 (formerly NICTA) · Foundational Machine-Learning & Statistical Research Lineage
The University of Queensland · Classifier & Quantitative Research Collaboration
Healthy Australia & Child Abuse Prevention Service (CAPS) · Clinical Instrument & Child-Protection Review Stewardship
Westpac · Supported Across Two Rounds of Westpac Social Scale Grant Funding
SAFE's development history includes foundational research contributions from NICTA, CSIRO/Data61 and The University of Queensland, clinical stewardship with Healthy Australia and the Child Abuse Prevention Service (CAPS), and two rounds of Westpac grant support. These historical research and funding contributions describe SAFE's development lineage and do not represent a third-party certification of every current product feature.
Australian Data Protection & Security Controls
Onshore AWS Sydney Core Hosting
Core child records, SAFE Assessments, Response notes and Expert Case Review records are hosted onshore in Australia on Amazon Web Services (AWS) in the Sydney Region (ap-southeast-2) across multiple Availability Zones. AWS Sydney infrastructure maintains independent ISO/IEC 27001, ISO/IEC 27017, ISO/IEC 27018 and SOC 1/2/3 certifications for cloud infrastructure under the AWS Shared Responsibility Model. Separate website measurement and non-content product analytics are governed by our Privacy Policy (/legal/privacy).
Encryption In Transit & At Rest
All web and API connections are encrypted in transit using HTTPS/TLS. Databases, file attachments and protected backups are encrypted at rest within isolated network environments in AWS Sydney.
Fine-Grained Access Privileges & Multi-Factor Authentication
Access to records is governed by fine-grained user privileges and strict organisation and service isolation, ensuring authorised staff only access records permitted for their specific service and privileges. Privileged administrative and reviewer access requires multi-factor authentication (MFA) and strict confidentiality controls.
Safeguarding-First Privacy, eSafety Alignment & Audit Trails
In alignment with the Online Safety Act 2021 (Cth) and eSafety Commissioner online safety standards, SAFE is restricted strictly to adult professionals (18+); children do not hold SAFE accounts, and sensitive child-protection and trauma records are ringfenced behind authenticated practitioner controls. System audit logs track authentication and record activity, and automated repository static analysis (DeepSource OWASP Top 10 and CWE/SANS Top 25 rulesets) supports secure software development. We never sell personal information.
Human-in-the-Loop Oversight, Anti-Bias Safeguards & Scientific Performance
Structured Observation
Authorised staff record structured Yes/No observations, Concern Ratings (1–5) and optional contextual Response notes.
Immediate Preliminary Guidance
Rules-based guidance (Act Now or No Urgent Flag) appears immediately on submission so urgent reporting never waits.
Deterministic Risk Triage
A Random Forest model evaluates structured assessment inputs to produce a Risk Probability that prioritises the review queue.
Human-in-the-Loop Expert Case Review
Child-protection specialists review every submitted assessment and record their professional review findings.
Intentional Input Boundaries, Human-in-the-Loop Architecture & Anti-Bias Design
SAFE's live Random Forest model operates strictly within a Human-in-the-Loop (HITL) decision-support architecture, reflecting Australian Institute of Family Studies (AIFS) best practice by pairing empirical actuarial consistency with clinical human judgement. The model evaluates a small, defined set of seven structured assessment inputs to help order records for human review. It never reads the words in free-text Response notes, attachments or offline clinical notes, and identifiable safeguarding records are never used to train public generative-AI models. As an intentional anti-bias and privacy safeguard, geographic location (postcode, suburb, coordinates), socio-economic indices (SEIFA), ethnicity, Aboriginal or Torres Strait Islander status, disability status, country of birth and language spoken at home are excluded from the model's inputs. SAFE provides upstream early identification and governance support; educators, service leaders and approved providers always remain responsible for safeguarding decisions and statutory reporting, and state and territory child-protection authorities remain responsible for statutory case management.
OCTOBER 2026 HISTORICAL CONCORDANCE REVIEW · >68,000 CHILD RECORDS
Accuracy
Overall agreement between model classification and recorded Human-in-the-Loop Expert Case Review verdict
Precision
Observed positive agreement in the evaluated cohort
Recall / Sensitivity
Identification of records classified At risk by Expert Case Review
F1 Score
Balanced measure combining precision and recall
These figures describe historical agreement between model classifications and recorded Expert Case Review verdicts across >68,000 child records in our October 2026 concordance review. They measure agreement with recorded expert verdicts in that cohort rather than independently confirmed abuse or a guarantee of future performance.
Service Availability & Live Status Monitoring
SAFE's public status monitor tracks the Web Application and Machine-Learning Service at five-minute intervals. Over the 90 days to 22 October 2026, each monitored endpoint recorded 100.00% observed availability (a dated monitoring observation for those endpoints rather than a future guarantee). Services should always continue urgent safeguarding and statutory reporting through their normal authority channels during any service interruption.
View current service status and uptime historyNational Principles, NQS QA2 & QA7 & Regulatory Alignment
SAFE provides structured assessment records, review timestamps and compliance reporting that help approved providers and service leaders evidence their safeguarding practice in alignment with the National Principles for Child Safe Organisations, the Education and Care Services National Law and National Regulations, and ACECQA National Quality Standard Quality Area 2 (Children's Health and Safety) and Quality Area 7 (Governance and Leadership), alongside the Privacy Act 1988 (Cth) and the Online Safety Act 2021 (Cth).
Controller–Custodian Governance, Incident Response & Record Retention
Approved providers remain the primary data controllers for their service records while SAFE acts as secure platform custodian. Access, correction and privacy requests (privacy@safe.inc) are assessed in coordination with the approved provider within a 30-day response target to protect child safety, mandatory-reporter confidentiality and third-party privacy under the Privacy Act 1988 (Cth). Incident-response procedures align with the OAIC Notifiable Data Breaches scheme. In accordance with the Education and Care Services National Regulations (reg 183, retention until age 25) and Royal Commission child sexual abuse recordkeeping standards (45 to 50 years), safeguarding records are preserved as required by law even after a child leaves a service or an account closes.
Enterprise, Government & Institutional Due Diligence
This page provides a public overview of SAFE's security architecture, privacy governance and scientific methodology. Comprehensive due-diligence documentation, including our mathematical specification and cohort evaluation paper (Technical Notes EN-01 to EN-04), application security and static-analysis reports, data-processing schedules, and privacy governance procedures, is available to prospective enterprise customers, government partners and qualified institutional reviewers under Mutual Non-Disclosure Agreement (MNDA).
Public policy resources include our Privacy Policy, Website Terms of Use and live service status page.
To request enterprise due-diligence documentation under NDA, or to report a suspected privacy or security vulnerability, contact privacy@safe.inc (please do not include child records, passwords or sensitive case material in an initial email).