Compliance

Residual risk

Level of harm remaining after mitigation controls.

How the laws define this

Both instruments agree that residual risk represents the risk or harm that remains after mitigation controls have been applied, serving as a baseline that determines escalation pathways and submission to a review committee. They diverge, however, in how specifically they categorize and act upon this risk. The NSW Artificial Intelligence Assessment Framework (AIAF) frames residual risk as the remaining level or likelihood of potential harm, noting that ratings determine if a project must be paused, halted, or submitted to the AI Review Committee. In contrast, DCS-2024-04 – Use of Artificial Intelligence by NSW Government Agencies defines residual risk as classifications produced by the AIAF, specifying that 'High' and 'Very High' ratings mandate escalation, additional assurance steps, and external review by the NSW AI Review Committee.

Synthesised from the 2 statutory definitions below. Each one is quoted in full, with its source.

How jurisdictions differ

What is distinctive in each definition, binding law in force first.

JurisdictionInstrumentStatusWhat is distinctive
New South Wales, AustraliaNSW Artificial Intelligence Assessment Framework (AIAF)In ForceAdds that residual risk ratings determine escalation pathways and whether a project must be paused, halted, or submitted to the AI Review Committee.
New South Wales, AustraliaDCS-2024-04 – Use of Artificial Intelligence by NSW Government AgenciesSupersededSpecifies that High and Very High residual risk ratings mandate escalation and external review by the NSW AI Review Committee, determining mandatory submission and additional assurance steps.

What it triggers

  • In the NSW Artificial Intelligence Assessment Framework (AIAF), residual risk ratings determine escalation pathways and whether a project must be paused, halted, or submitted to the AI Review Committee.source
  • Under DCS-2024-04 – Use of Artificial Intelligence by NSW Government Agencies, High and Very High residual risk ratings required escalation and external review by the NSW AI Review Committee until it was superseded.source

Definitions (2)

The level of potential harm or the likelihood of harm that remains after mitigation controls have been applied; residual risk ratings determine escalation pathways and whether a project must be paused, halted, or submitted to the AI Review Committee.

Risk classifications produced by the AIAF indicating remaining (residual) risk after mitigations; 'High' and 'Very High' residual risk ratings mandate escalation and external review by the NSW AI Review Committee. These classifications determine mandatory submission and additional assurance steps.

Not to be confused with

High/Very High residual risk

High/Very High residual risk refers specifically to the residual risk ratings that trigger external expert review, while residual risk is the broader level of remaining harm or likelihood after mitigations.

Questions and answers

What happens if residual risk is high under NSW AIAF?

Residual risk ratings determine escalation pathways and whether a project must be paused, halted, or submitted to the AI Review Committee. This applies after mitigations have been applied.

What did high residual risk require under the superseded DCS-2024-04?

High and Very High residual risk ratings required escalation and external review by the NSW AI Review Committee. These classifications determined mandatory submission and additional assurance steps until the instrument was superseded.

How does residual risk affect AI projects in NSW government?

Residual risk is the level of potential harm or likelihood remaining after mitigations, with ratings triggering specific actions like escalation or review. High or Very High ratings lead to mandatory external review processes.