Governing with artificial intelligence in the public interest

The famous Uruguayan writer Mario Benedetti once observed that “when we thought we?had the answers, the questions suddenly changed.” This is what is happening to governments around the world that are grappling to unlock the potential of artificial intelligence (AI), while mitigating its risks. They are considering a range of regulatory approaches with the aim of governing AI rather than being governed by it. In the UK, the new Labour government is fast-forwarding its strategy to leverage AI for reimagining the state and modernising public services. A number of countries, such as Norway, are creating dedicated ministries to steer AI from the centre.

Delivering better and smarter government?

As regulators, governments have a central role in steering and shaping the responsible development of AI in the public interest. As users, they must also ensure that they deploy AI solutions within the public sector strategically to modernise government and improve services. The use cases of AI across core government functions are expanding rapidly, including in sensitive policy domains such as tax fraud, law enforcement, justice administration and welfare benefits.

For governments around the world, governing with AI is becoming as important as governing AI. Deploying AI in a strategic, yet responsible manner requires steering it from the centre with strategic intent and the necessary guardrails. This is partly driven by recent high-profile incidents, for example in the Netherlands and Australia, as documented by the OECD AI Observatory. Similarly, the EU AI Act classifies most uses of AI by government as “high-risk” requiring heightened scrutiny, in particular in areas of facial recognition, law enforcement and judicial administration.

The potential of AI is indeed immense but its deployment has tended to be fragmented in ad-hoc pilots that seldom scale for impact. This is changing though, and rapidly. Under the Italian presidency, the G7 has started to address the specific challenges of AI in the public sector. In its ministerial meeting on digital transformation in Cuomo, Italy, on 15 October 2024 it adopted a toolkit to guide AI deployment in the public sector, prepared by the OECD and UNESCO.

Our recent OECD policy paper on Governing with Artificial Intelligence: Are Governments Ready? sets a framework to help governments leverage AI for impact, in line with the principles of OECD Recommendation on AI. AI can help transform government in four main ways:

·????? Enhancing effectiveness. AI can help design better policies and target public spending, with more granular insights and predictive analytics. For example, Korea uses AI to forecast emerging infectious disease and develop policy responses.

·????? Improving efficiency. AI can increase the internal efficiency of government operations. AI systems can handle large volumes of data, identify complex patterns, and adapt and learn from new data to improve their accuracy. For example, the Canadian Revenue Agency has introduced AI-powered chatbots to help people navigate complex information and procedures to facilitate tax obligations.

·????? Enhancing responsiveness – AI can enable more personalised, proactive and integrated government services centred on people needs and their life events. For example, Finland is using AI to increase the take-up of unemployment assistance and minimum income benefits. Costa Rica has built-in an algorithm to identify people in poverty and the social benefits they are entitled to.

·????? Strengthening integrity – AI has become a powerful ally of transparency and oversight around the world, helping raise red-flags in government procurement, tax obligations, and welfare benefits. Government auditors and procurement agencies have been particularly active in deploying AI against corruption risks. For example, the Brazilian federal auditor developed Alice, a tool that automates procedures and enables continuous and large-scale auditing of public tenders.

?

Ensuring strategic and responsible adoption

Governing with AI requires governing AI in the public sector. To realise the full potential of AI, governments need to enhance its enablers and establish the necessary guardrails to ensure its strategic and responsible use.

·????? Strategy – Setting a whole-of-government approach with common standards and policy guidelines to support the adoption of AI across the public sector. In 2023, the US issued an executive order to ensure the safe, secure and trustworthy use of AI across the federal government, setting in motion a government-wide effort at better governing AI under the steering of the White House’s Office of Management and Budget.

·????? Enablers – AI in government requires solid digital public infrastructure in particular robust data governance (including data ethics), digital identity and digital payments. It also needs adequate digital skills and smart procurement rules. For example, the UK has developed ethical guidelines for government contracting of AI. By setting rules for procuring AI, governments can advance the public interest but also tilt the market. Lastly, novel partnerships with agile govtech startups can also facilitate uptake of AI in the public sector.

·????? Guardrails – Transparency standards and effective oversight of public algorithms are critical for public trust. The complexity of AI systems often makes it difficult to understand how decisions are made, challenging accountability and traceability. Citizens should be able to understand the operation and use of algorithms and know where to turn to with questions. Governments (France and Chile) and cities (Amsterdam, New York, Helsinki) are establishing open registers of public algorithms. Spain has established a dedicated agency for AI supervision. Oversight institutions such as ombudspersons institutions, data protection authorities, and audit agencies such as in the UK (NAO), the US (GAO) and the Netherlands (Rekenkamer, NL), are stepping-up their scrutiny of AI use in government.

The digital future that we want

Getting the governance of AI right is a defining challenge of our digital times. It is not about the technology itself, but how to best steer its use in the public interest and for the common good, in line with our own principles, values and rights. As Mahatma Gandhi reminded us, “the future depends on what we do in the present.” France’s AI Action Summit in early 2025 will provide another opportunity to advance the responsible use of AI in government.