|
| 1 | +--- |
| 2 | +layout: page |
| 3 | +title: "AI Glossary (Australia)" |
| 4 | +description: "A practical glossary of key AI terms in the Australian context. Comprehensive definitions for AI safety, governance, and compliance in Australia." |
| 5 | +keywords: "AI glossary Australia, AI terms, AI definitions, AI safety glossary, AI governance terms, Australian AI terminology, AI compliance glossary" |
| 6 | +author: "SafeAI-Aus" |
| 7 | +robots: "index, follow" |
| 8 | +og_title: "AI Glossary (Australia)" |
| 9 | +og_description: "A practical glossary of key AI terms in the Australian context" |
| 10 | +og_type: "article" |
| 11 | +og_url: "https://safeaiaus.org/glossary/" |
| 12 | +og_image: "assets/safeaiaus-logo-600px.png" |
| 13 | +twitter_card: "summary_large_image" |
| 14 | +twitter_title: "AI Glossary (Australia)" |
| 15 | +twitter_description: "A practical glossary of key AI terms in the Australian context" |
| 16 | +canonical_url: "https://safeaiaus.org/glossary/" |
| 17 | +--- |
| 18 | + |
| 19 | +# AI Glossary (Australia) |
| 20 | + |
| 21 | +*A practical glossary of key AI terms in the Australian context. Provided by SafeAI Aus (safeaiaus.org) under CC BY 4.0 licence.* |
| 22 | + |
| 23 | +--- |
| 24 | + |
| 25 | +## Artificial Intelligence (AI) |
| 26 | +The simulation of human intelligence by machines, especially computer systems. In Australia, AI use is guided by government guardrails, privacy law and international standards such as ISO/IEC 42001. |
| 27 | + |
| 28 | +## AI Guardrails (Australia) |
| 29 | +Voluntary principles from the Australian Government that set practical expectations for safe and responsible AI. They help organisations align practice with community standards and emerging regulation. |
| 30 | + |
| 31 | +## AI Incident |
| 32 | +Any harmful, biased, unsafe or unexpected behaviour from an AI system (e.g., discriminatory outputs, privacy breaches, safety risks). Log and review with an [AI Incident Report](ai-incident-report-form.md). |
| 33 | + |
| 34 | +## AI Inventory (System Register) |
| 35 | +A catalogue of AI systems and use cases across the organisation, including owners, purpose, data sources, risks and status. Supports accountability and audits. |
| 36 | + |
| 37 | +## AI Risk Assessment |
| 38 | +A structured process to identify, analyse and mitigate risks for AI systems and use cases. Often mapped to ISO/IEC 42001 (management systems) or ISO/IEC 23894 (risk management). Start with the [AI Risk Assessment Template](ai-risk-assessment-checklist.md). |
| 39 | + |
| 40 | +## AI Use Policy |
| 41 | +An internal policy that sets boundaries, roles and responsibilities for AI tools. Defines acceptable use, prohibited use, data handling and incident escalation. See the [AI Use Policy Template](ai-use-policy.md). |
| 42 | + |
| 43 | +## Algorithmic Impact Assessment (AIA) |
| 44 | +An assessment of potential impacts (e.g., fairness, safety, human rights). In Australia this is often paired with a Privacy Impact Assessment (PIA) where personal information is processed. |
| 45 | + |
| 46 | +## Anonymisation / De-identification |
| 47 | +Techniques to reduce the risk of re-identification in datasets. Australia commonly uses “de-identification” under the Privacy Act; truly irreversible anonymisation is difficult in practice. |
| 48 | + |
| 49 | +## APPs (Australian Privacy Principles) |
| 50 | +The 13 principles under the **Privacy Act 1988 (Cth)** that govern handling of personal information by APP entities (most Australian Government agencies and many businesses). |
| 51 | + |
| 52 | +## ASD Essential Eight |
| 53 | +Australian Signals Directorate’s baseline mitigation strategies for cyber security (e.g., patching, MFA). Not AI-specific but essential for AI system hardening and data protection. |
| 54 | + |
| 55 | +## Bias (Algorithmic Bias) |
| 56 | +Systematic errors that favour or disadvantage groups (e.g., by gender, ethnicity, age). Manage via representative data, fairness testing, documentation and human oversight. |
| 57 | + |
| 58 | +## C2PA (Content Authenticity) |
| 59 | +An open standard for attaching provenance metadata to content. Useful for signalling AI-generated or edited media and supporting authenticity claims. |
| 60 | + |
| 61 | +## Consumer Data Right (CDR) |
| 62 | +Australian framework enabling data portability in designated sectors. Relevant when AI uses consumer data that may be shared or accessed under CDR rules. |
| 63 | + |
| 64 | +## Content Moderation / Safety Filters |
| 65 | +Controls that reduce harmful or disallowed outputs (e.g., hate speech, self-harm). Often combined with human review for higher-risk contexts. |
| 66 | + |
| 67 | +## Continuous Improvement (AI) |
| 68 | +Ongoing monitoring, feedback and updates to models, prompts and controls. A core requirement in management-system approaches such as ISO/IEC 42001. |
| 69 | + |
| 70 | +## Data Residency & Sovereignty |
| 71 | +Where data is stored and which laws apply. Australian organisations often prefer Australian regions for regulated datasets and clarity on cross-border transfers. |
| 72 | + |
| 73 | +## Data Source Register |
| 74 | +A record of datasets used for training, fine-tuning or RAG. Includes lineage, licences, sensitivity and quality notes. |
| 75 | + |
| 76 | +## Drift (Data/Model Drift) |
| 77 | +Performance degradation when real-world data changes away from training assumptions. Detect via monitoring, evals and periodic re-training. |
| 78 | + |
| 79 | +## Evals (Evaluation) |
| 80 | +Systematic tests for quality, safety and robustness (e.g., accuracy, bias, jailbreak resistance). Should be repeatable and linked to risk level. |
| 81 | + |
| 82 | +## Explainability (XAI) |
| 83 | +The ability to understand or describe how an AI system produced an output. Methods range from interpretable models to post-hoc explanations and model cards. |
| 84 | + |
| 85 | +## Fine-Tuning |
| 86 | +Adapting a model on domain-specific data to improve performance. Requires careful governance of data rights, privacy and overfitting risks. |
| 87 | + |
| 88 | +## Foundation Model |
| 89 | +A large pre-trained model (text, image, multimodal) adaptable to many tasks (e.g., via prompting, fine-tuning or RAG). |
| 90 | + |
| 91 | +## Hallucination |
| 92 | +A confident but incorrect output from a generative model. Mitigate with retrieval augmentation, constraints, verification and human review. |
| 93 | + |
| 94 | +## Human-in-the-Loop (HITL) |
| 95 | +Design pattern where humans review or approve AI outputs for higher-risk tasks, or provide feedback to improve models. |
| 96 | + |
| 97 | +## ISO/IEC 23894 (AI Risk Management) |
| 98 | +International guidance for managing AI risks across the lifecycle. Complements management-system standards and local guardrails. |
| 99 | + |
| 100 | +## ISO/IEC 42001 (AI Management System) |
| 101 | +International standard for governing AI (policy, risk, controls, monitoring and continual improvement). Useful for phased adoption by SMEs. |
| 102 | + |
| 103 | +## Large Language Model (LLM) |
| 104 | +A model trained on very large text corpora to generate and understand language. Examples include ChatGPT, Claude and Gemini. |
| 105 | + |
| 106 | +## Logging & Auditability |
| 107 | +Recording inputs, outputs, decisions and system changes so incidents can be investigated and controls verified. |
| 108 | + |
| 109 | +## Model Card / Data Card |
| 110 | +Documentation describing a model or dataset: purpose, training data, limitations, risks, metrics and intended use. |
| 111 | + |
| 112 | +## Model Monitoring |
| 113 | +Operational tracking of model quality, drift, latency, cost and safety incidents in production. |
| 114 | + |
| 115 | +## NDB Scheme (Notifiable Data Breaches) |
| 116 | +Australian scheme requiring notification to the OAIC and affected individuals when an eligible data breach is likely to cause serious harm. |
| 117 | + |
| 118 | +## NLP (Natural Language Processing) |
| 119 | +AI methods that process and generate human language. Powers chatbots, search, summarisation and translation. |
| 120 | + |
| 121 | +## OAIC (Office of the Australian Information Commissioner) |
| 122 | +Australia’s independent regulator for privacy and information access. Oversees the Privacy Act and NDB scheme. |
| 123 | + |
| 124 | +## Open-Weight vs Closed-Weight Models |
| 125 | +Open-weight models allow running the model locally or in private environments; closed-weight models are accessed via APIs. Each has different governance and risk profiles. |
| 126 | + |
| 127 | +## Personal Information (Australia) |
| 128 | +Information or an opinion about an identifiable individual, as defined in the **Privacy Act 1988 (Cth)**. Includes obvious identifiers and information that could reasonably identify someone. |
| 129 | + |
| 130 | +## PIA (Privacy Impact Assessment) |
| 131 | +Analyses privacy impacts and mitigations for projects that handle personal information. Often required or strongly recommended for AI implementations in Australia. |
| 132 | + |
| 133 | +## Prompt / Prompt Engineering |
| 134 | +The inputs, instructions and context provided to a generative model. Good prompt design improves reliability, safety and usefulness. |
| 135 | + |
| 136 | +## Prompt Injection / Jailbreak |
| 137 | +Adversarial inputs that subvert a model’s instructions or controls. Mitigate with input filtering, retrieval isolation, guardrails and least-privilege design. |
| 138 | + |
| 139 | +## Provenance / Watermarking |
| 140 | +Signals that indicate how content was created (e.g., AI-generated) or where it came from. Supports trust, authenticity and moderation workflows. |
| 141 | + |
| 142 | +## RAG (Retrieval-Augmented Generation) |
| 143 | +Architecture where the model retrieves trusted, up-to-date documents and uses them as context before generating an answer. Reduces hallucinations and improves domain accuracy. |
| 144 | + |
| 145 | +## RBAC (Role-Based Access Control) |
| 146 | +Security control restricting who can run models, access prompts, view logs or export data. |
| 147 | + |
| 148 | +## Responsible AI |
| 149 | +Practices ensuring AI systems are safe, ethical, transparent and aligned to human values. In Australia, anchored by government guardrails, privacy law and recognised standards. |
| 150 | + |
| 151 | +## Risk Appetite / Tolerance |
| 152 | +The level and types of risk an organisation is willing to accept in AI adoption. Guides control strength, oversight and rollout pace. |
| 153 | + |
| 154 | +## Safety Case (AI) |
| 155 | +Documented argument and evidence that an AI system is acceptably safe for its intended use. More common for higher-risk sectors. |
| 156 | + |
| 157 | +## Sensitive Information (Australia) |
| 158 | +A subset of personal information (e.g., health, biometrics, racial or ethnic origin, religious beliefs, sexual orientation) that attracts higher protections under the Privacy Act. |
| 159 | + |
| 160 | +## Shadow AI |
| 161 | +Use of unapproved AI tools by staff. Mitigate through clear policy, training, approved toolkits and monitoring. |
| 162 | + |
| 163 | +## Template Library |
| 164 | +A collection of standardised documents (policy, risk, incident, vendor, register) to support safe AI adoption. See the [Template Library](policy-template-library.md). |
| 165 | + |
| 166 | +## Transparency |
| 167 | +Clarity about how an AI system works, what data it uses and how decisions are made. Supports accountability, trust and compliance. |
| 168 | + |
| 169 | +## Vendor Due Diligence (AI) |
| 170 | +Assessing third-party AI tools for security, privacy, reliability and compliance. Use the [AI Vendor Evaluation Checklist](ai-vendor-evaluation-checklist.md). |
| 171 | + |
| 172 | +--- |
| 173 | + |
| 174 | +# 🔗 Related Resources |
| 175 | +- [Template Library](policy-template-library.md) |
| 176 | +- [AI Safety Standards (Australia & International)](voluntary-ai-safety-standard-10-guardrails.md) |
| 177 | +- [AI Project Register](ai-project-register.md) |
| 178 | +- [AI Use Policy Template](ai-use-policy.md) |
| 179 | +- [AI Risk Assessment Template](ai-risk-assessment-checklist.md) |
| 180 | +- [AI Incident Report Form](ai-incident-report-form.md) |
| 181 | + |
| 182 | +--- |
| 183 | + |
| 184 | +## Attribution |
| 185 | +This glossary is published by **SafeAI Aus** under [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/). You may copy, adapt and reuse with attribution: *“Source: SafeAI Aus (safeaiaus.org)”*. |
| 186 | + |
| 187 | +--- |
| 188 | + |
| 189 | +<!-- JSON-LD: Glossary schema for crawlers and LLMs --> |
| 190 | +<script type="application/ld+json"> |
| 191 | +{ |
| 192 | + "@context": "https://schema.org", |
| 193 | + "@type": "DefinedTermSet", |
| 194 | + "name": "AI Glossary (Australia)", |
| 195 | + "inLanguage": "en-AU", |
| 196 | + "url": "https://safeaiaus.org/glossary", |
| 197 | + "creator": { "@type": "Organization", "name": "SafeAI Aus", "url": "https://safeaiaus.org" }, |
| 198 | + "license": "https://creativecommons.org/licenses/by/4.0/" |
| 199 | +} |
| 200 | +</script> |
0 commit comments