AI governance for government and the public sector.

EU AI Act

AI used in law enforcement, migration and border control, administration of justice, and critical infrastructure is classified as high-risk or subject to outright prohibition. AI-powered biometric surveillance in public spaces is prohibited with narrow exceptions. Public sector agencies are among the most heavily regulated AI deployers.

Administrative law and procedural fairness

Government decisions affecting individuals must comply with procedural fairness obligations — including where AI informs or makes those decisions. Individuals typically have rights to reasons, to challenge decisions, and to have challenges reviewed by humans capable of independent judgment.

Data protection and privacy

Government agencies process sensitive personal data at scale. AI that processes this data must comply with applicable privacy law, which in many jurisdictions imposes stricter obligations on government entities than on private organisations.

Public sector procurement rules

Government procurement of AI systems is governed by procurement law that creates transparency, accountability, and value-for-money obligations. Sole-source procurement of AI systems without adequate due diligence creates governance and procurement law risk.

Robodebt (Australia)

The Australian government's automated debt assessment program raised debts against approximately 443,000 welfare recipients using an income averaging methodology that was unlawful. The Robodebt Royal Commission documented governance failures at every level — from legal review of the automated methodology, to escalation of concerns, to the treatment of complaints as a system management problem rather than a governance signal.

Automated benefits decisions without human review

Multiple jurisdictions have encountered legal challenges to AI-assisted government benefit decisions where human review was nominal rather than genuine. Courts have increasingly required that human reviewers be capable of independent judgment — not merely capable of approving AI outputs.

Algorithmic bias in public services

AI systems used in public services — housing allocation, child protection risk assessment, policing — have been found to produce systematically different outcomes for minority communities. In government contexts, this creates both legal exposure and democratic legitimacy concerns.

Inadequate transparency about AI use

Citizens often do not know when AI is being used in decisions about them. The EU AI Act requires disclosure for high-risk AI in government contexts. Proactive algorithmic impact assessments and public registers of AI systems are becoming a democratic accountability expectation.

Do you maintain a public register or internal inventory of AI systems used in decisions affecting citizens, with their risk classification?

For AI systems that inform decisions with legal or significant effects on individuals, what is the human review process — and are reviewers capable of genuinely independent judgment?

Have you conducted algorithmic impact assessments for high-impact AI systems, and are those assessments publicly available?

What are your regulatory notification obligations if an AI system causes serious harm — and do you have a process to identify and report qualifying incidents?

Has your AI procurement been reviewed against applicable procurement law for transparency, accountability, and value-for-money obligations?

How does parliamentary or legislative oversight apply to your AI deployments — and what information is provided to elected representatives on request?