AU Enterprise AI Glossary
24 governance, risk, regulatory, and technical terms — the vocabulary behind the AEAI methodology and regulatory tracker, free to reference and link to.
Governance
5AI Governance Framework
The documented policies, accountabilities, and review processes an organisation uses to deploy AI responsibly.
AI Model Register
A maintained inventory of every AI system in use, including owner, purpose, data inputs, and risk rating.
Chain of Custody (AI Audit Trail)
An immutable, verifiable record of how a score or decision was produced, from raw input to final output.
Human-in-the-Loop (HITL)
A control design where a human must review or approve an AI-generated decision before it takes effect.
Responsible AI
An umbrella term for practices that ensure AI systems are fair, safe, transparent, and accountable.
Risk
6AI Incident
An event where an AI system causes harm, fails unexpectedly, or produces a materially wrong outcome.
AI Red Teaming
Adversarial testing of an AI system to surface failure modes, jailbreaks, and security weaknesses before deployment.
AI Supply Chain Risk
Risk introduced by third-party models, datasets, or vendors an organisation relies on but does not directly control.
Bias Audit
A structured review testing whether an AI system produces systematically unequal outcomes across groups.
Model Risk Management (MRM)
The discipline of identifying, measuring, and controlling the risk that a model behaves incorrectly or is misused.
Shadow AI
AI tools used inside an organisation without IT or governance approval — the AI-era version of shadow IT.
Regulatory
3Data Sovereignty
The principle that data is subject to the laws of the country in which it is collected or stored.
Notifiable Data Breach (NDB)
The OAIC scheme requiring organisations to report data breaches likely to cause serious harm.
Voluntary AI Safety Standard (VAS)
DISR's 10 voluntary guardrails for responsible AI, published September 2024, the de facto AU governance baseline.
Technical
9Agentic AI
AI systems that autonomously plan and execute multi-step tasks, rather than responding to a single prompt.
AI Gateway
A managed proxy layer that routes, logs, and governs all AI model calls across an organisation.
Explainability (XAI)
The degree to which an AI system's decisions can be understood and justified by a human reviewer.
Fine-tuning
Further training a pre-trained foundation model on a smaller, domain-specific dataset to specialise its behaviour.
Foundation Model
A large, general-purpose AI model trained on broad data, adaptable to many downstream tasks via prompting or fine-tuning.
Guardrails
Automated constraints that filter, block, or modify AI inputs and outputs to prevent unsafe or unwanted behaviour.
Hallucination
When an AI model generates confident, fluent output that is factually incorrect or entirely fabricated.
Prompt Injection
An attack that manipulates an AI system's behaviour by embedding malicious instructions in its input.
Retrieval-Augmented Generation (RAG)
An architecture that grounds a model's answers in retrieved documents rather than relying on memorised training data.