New approaches to anticipatory regulation: the Regulatory Innovation Office (RIO) in the UK and the new AI regulators

In short

New regulatory bodies like the UK’s Regulatory Innovation Office signal a growing recognition that institutions must adapt to govern emerging technologies. But while their mandates are future-facing, many still follow traditional structures that may limit their effectiveness. The real opportunity lies in adopting new institutional design principles — like intelligence as a core function and outside-in approaches — to make regulation as dynamic as the challenges it seeks to address.

An important new field for public institutional design is regulation – with new kinds of regulator being established for fields including AI where the pace of change is fast, and in the UK a new Regulatory Innovation Office which started work in 2025 succeeding a previous ‘Regulatory Pioneers Fund’ that had supported regulators to adopt new methods.

These new cross-cutting institutions aim to help the governance of fast-changing and complex technologies that bring both major opportunities as well as risks.

Background

The background is the challenge of how law and regulation should cope with fast-changing technologies and industries, balancing the risks that come with new ideas and the risks of crushing them.

In recent decades, the dominant ideas about regulation emphasised that it should be constant, simple and predictable. If it were, markets could do what they do best, finding ways to optimise the implementation of new technologies and ideas. A host of regulatory theory and regulatory institutions emerged from the 1970s onwards to put the economic theory of regulation into practice. It adapted long-standing ideas about market failure, externalities and the risks of regulatory capture. It tended to emphasise the role of regulators in promoting competition and contestability more than a previous generation of ideas that were mainly about constraining the abuses of monopolists.

Its advocates promised to replace the capricious decisions of bureaucrats with more arm’s-length rule-makers and more rational rules. The promised result would be more competition and a better deal for consumers. The idea was that regulators should not try to second-guess the direction of technological change. Instead, they should set the game’s rules and then stand back.

Challenges

Traditional regulatory theory can still work fairly well for stable industries with relatively stable technologies. However, it struggles to cope with more fluid, dynamic, and uncertain fields, particularly ones where the boundaries between industries are constantly changing. This is particularly the case with many of the new platform-based business models that have come to dominate so many industries, and with the extraordinary pace of developments in AI, particularly since the emergence of Large Language Models.

Regulators have always faced an inescapable dilemma on timing. Acting to regulate too early can kill off or freeze innovative business models with a potential for public good. Acting too late can leave consumers exposed to harm or allow new monopolies to become entrenched.

In the past, regulators assumed that they could ignore new developments until they reached a certain scale. Likewise, new firms didn’t engage with regulators until they hit a large enough scale. But speed undermines both sets of assumptions. Small firms can become big very fast, much faster than in a predominantly material economy, and that’s forcing attention to quick and lean ways of linking what may be a large pool of potential new approaches and innovators to the limited resources of regulators.

The related challenge is that the structure of markets has changed radically with new forms of market power embodied in platforms. These tend to form quasi-monopolies, giving great gains to consumers in terms of lower prices but also potential harms. Market dynamics can risk reducing competition and innovation as large incumbents are incentivised to buy out, and often shut down, potentially threatening incomers.

The rise of anticipatory regulation

Governments around the world are grappling with these questions. A family of new methods loosely as ‘anticipatory regulation’ is now emerging. These recast regulations to assist in the emergence of new technological tools while also allowing faster responses to ensure that the public isn’t exploited and that new dangers are averted.

These include:

much more dialogue between innovators and regulators
the use of sandboxes
testbeds
contingent licensing, which allows an innovator to test out a new approach in the real world
the use of open data (pioneered in banking)
a constant push to remove unnecessary or obsolete regulations

Elements of this aren’t new. Smart governments have long tried to achieve a better alignment of technology development, market regulation and public policy, pre-empting significant shifts, as Scandinavian governments successfully did with GSM mobile a generation ago. There are some good recent examples — in fields such as human fertilisation — of imaginative exercises to address the intertwined issues of ethics, public confidence and technology development in emerging fields, deliberately anticipating potential problems.

Culling unnecessary regulations

And there have been many good examples of cutting regulations: from Portugal’s SIMPLEX programme to the UK Red Tape Challenge, which crowd-sourced ideas on culling useless rules; and from South Korea’s ‘Regulatory Guillotine’ in the late 1990s to waves of regulation culling in Australia and New Zealand at a similar time.

Some of the best examples are ration rules and laws, like a diet or a ‘bureaucratic budget’. Canada, for example, has followed a ‘one for one’ approach — one rule cut for any new one introduced. And in the 2000s, the Netherlands targeted a 25% annual cut in red tape for business (later copied in other countries).

Governments can also aim to lighten the cognitive load on citizens when designing any policies (Australia’s Treasury has long been a good example of this, the UK a particularly bad one, with ever more complex tax and benefits rules that work on paper but make life difficult for citizens and businesses). ‘Once only’ digital models are also a good way to cut cognitive loads, so that you only have to tell the government once when, for example, you move home or someone dies (though experience over the last 20 years shows that they’re quite difficult to implement).

For regulators, however, it continues to be a challenge to navigate economies in which:

the pace of change is rapid, as Moore’s law continues to operate;
barriers to entry are often very low;
ability to operate across sectoral boundaries is high (with technologies like 3D printing, platforms and much AI);
business models encourage big firms to operate across multiple sectors (as Alphabet now does) or run whole ecosystems of products and services (like Apple or Amazon);
market dynamics depend much more on data and algorithms than the traditional tools for asserting power;
new capabilities in machine intelligence are throwing up a host of legal, ethical and practical challenges, which traditional regulatory and legal concepts simply don’t fit.

New institutions

The arrival of AI has belatedly prompted the design of new regulators. Early thinking on this began over ten years ago. Across the EU, national governments are set to create some institutions to implement their new AI law, with Spain’s AESIA a first example, while China set up a powerful Cyberspace Administration in the early 2010s, which has now extended its role to AI.

While the UK has not followed this route yet, it did respond to the broader anticipatory regulation agenda by creating an innovation fund to help regulators adapt (titled ‘Regulatory Pioneers Fund’, launched in 2018). Nesta was involved in many of the resulting projects, for example, working on the use of AI in law and on drones

In 2024, the new Labour government announced a ‘Regulatory Innovation Office’, building on this work and incorporating the RPF. It began work in 2025 and will be an interesting experiment to watch.

In each case, these remain quite traditional in structure and form, despite their forward-looking remits. Looking to the future, it will be interesting to see how any of these adopt some of the new design principles (set out in our Playbook and elsewhere), including the focus on intelligence as a core function, mesh models of organisation, and outside-in design approaches.

White paper #002: A Well-Architected Framework for public procurement

Software that does not constantly evolve is outcompeted, vulnerable to threats, and abandoned by users. The same can be true for institutions.

Playbook: Designing new institutions and renewing existing ones

The UNDP-TIAL-Demos Helsinki Playbook helps teams have a conversation about strategy and tactics for renewing existing or designing new institutions.

Designing new public institutions for the UK in the 2020s and beyond

This paper looks at the history of institutional design and the many potential priorities for creating new institutions in the UK over the next decade