Key terms
For definitions of existing terms, refer to key terms in the Technical standard for government's use of artificial intelligence: key terms.
Control tower
Centralised governance that monitors, manages, and provides oversight of AI system operations, including performance, risks, costs, and compliance.
Data exfiltration
The theft or unauthorised removal or movement of any data from a device or system. This typically involves stealing data through various cyberattack methods.
Human-in-the-loop (HITL)
Refers to systems where a human actively participates in the decision-making process. In HITL, a human will review inputs and outputs and make the final decision.
Human-on-the-loop (HOTL)
Allows the system to operate autonomously while supervised by a human who holistically monitors the system and intervenes when needed.
Human-out-of-the-loop (HOOTL)
Operates fully autonomously and makes decisions independently without human interaction.
Human oversight
Governance mechanisms that ensure human involvement in supervising, approving, or auditing AI actions to maintain accountability and safety.
Observability
Capability to monitor, trace, and analyse the behaviour of agents and AI systems through logs, metrics, and audit trails to ensure transparency, detect issues, support debugging, and enable governance and compliance oversight.