Compliance

assurance

Evidence-based confidence an AI system is safe, lawful and fit for purpose.

How the laws define this

Regulatory standards converge on the principle that assurance requires demonstrating that an artificial intelligence technology operates lawfully, safely, and reliably in accordance with applicable rules and policies. Both the National Framework for the Assurance of Artificial Intelligence in Government and the Defence Artificial Intelligence Strategy frame assurance as a continuous lifecycle process relying on testing, documentation, monitoring, and governance. In contrast, National Security Presidential Memorandum/NSPM-11 — Artificial Intelligence in the National Security Enterprise diverges by defining assurance as a mandatory design requirement focused on specific system safeguards, stipulating that adopted technologies must be designed to be robust, steerable, controllable, and protected from unauthorized modification or disablement. The procedural focus also varies among the instruments: the Defence Artificial Intelligence Strategy explicitly requires validation, verification, risk assessment, and staged sign-offs prior to operational use, whereas the National Framework for the Assurance of Artificial Intelligence in Government highlights retaining evidence to give decision-makers and the public confidence in system outcomes.

Synthesised from the 3 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.

Jurisdiction	Instrument	Status	What is distinctive
United States	National Security Presidential Memorandum/NSPM-11 — Artificial Intelligence in the National Security Enterprise	In Force	Mandates reliability, robustness, steerability, controllability, legal/policy compliance, and protection from unauthorized modification or disablement specifically for AI in US national security.
Australia	National Framework for the Assurance of Artificial Intelligence in Government	In Force	Emphasizes provision and retention of evidence and processes like testing, documentation, monitoring, and governance to build confidence for AI systems in Australian government across their lifecycle.
United Kingdom	Defence Artificial Intelligence Strategy	In Force	Specifies validation, verification, risk assessment, and staged sign-offs prior to operational use to demonstrate safety, reliability, and legal/ethical compliance in UK Defence AI.

Definitions (3)

Assurance means the provision and retention of evidence and processes that demonstrate an AI system is safe, lawful and fit for purpose across its lifecycle. It encompasses testing, documentation, monitoring, governance and other practices that give decision‑makers and the public confidence in system behaviour and outcomes.

National Framework for the Assurance of Artificial Intelligence in GovernmentDefinition 1 of 3

Processes, testing, governance and documentation across the AI lifecycle required to demonstrate safety, reliability and legal/ethical compliance, including validation, verification, risk assessment and staged sign-offs prior to operational use.

Defence Artificial Intelligence StrategyDefinition 2 of 3

A critical pillar of NSPM-11, Assurance mandates that all adopted AI technologies are designed to be reliable, robust, steerable, and controllable, operating in accordance with applicable laws and government policies, and protected from unauthorized modification or disablement.

National Security Presidential Memorandum/NSPM-11 — Artificial Intelligence in the National Security EnterpriseDefinition 3 of 3

Not to be confused with

Assurance model

Assurance model refers to a structured, risk-proportionate framework for AI assurance, while assurance is the broader processes and evidence itself.

third‑party AI assurance

Third-party AI assurance is independent evaluation by external organisations, whereas assurance encompasses all processes including internal testing and governance.

AI assurance ecosystem

AI assurance ecosystem is the coordinated mechanisms for auditing AI, while assurance is the individual processes, testing, and documentation applied to specific systems.

Questions and answers

What does AI assurance require under NSPM-11?

Under NSPM-11, assurance mandates that adopted AI technologies for national security are reliable, robust, steerable, controllable, compliant with laws and policies, and protected from unauthorized changes. It is a critical pillar for AI in US defense and intelligence.

How is assurance defined in Australia's National Framework for AI in Government?

Assurance involves providing and retaining evidence and processes like testing, documentation, monitoring, and governance to show AI systems are safe, lawful, and fit for purpose throughout their lifecycle. This builds confidence for decision-makers and the public.

What processes does the UK Defence AI Strategy include in assurance?

The strategy defines assurance as processes, testing, governance, and documentation across the AI lifecycle, including validation, verification, risk assessment, and staged sign-offs before operational use to ensure safety, reliability, and compliance.