The UK’s cautious path on AI

Artificial intelligence (AI) is increasingly seen as one of the most influential milestones of human progress, comparable to the invention of the printing press, the introduction of manufacturing assembly lines, or the internet. Increasingly,  countries around the world are looking at AI as a crucial ingredient for their future prosperity, and this influences their policies and regulatory approaches towards this technology. 

In the UK, successive governments have been extremely cautious about taking regulatory steps that could jeopardise AI’s contribution to growth, but their hesitation has also created uncertainty among businesses and citizens about how the nation balances AI’s opportunities and risk.

An AI-friendly regulatory landscape

Recently, the UK published a proposal for a UK AI Growth Lab. Through this proposal, the Government seeks to implement a sandbox-based model to ensure that existing legislation keep pace with rapid AI development while still enabling effective regulatory oversight. The use of sandboxes are not new in the digital regulatory space. 

For instance, the UK Information Commissioner’s Office has actively supported this concept as a mechanism to test novel data processing practices, and the use of sandboxes in the context of AI is meant to be a cornerstone of the EU’s AI Act. However, what makes the UK approach particularly unique is that the proposed sandboxing strategy would likely lead to the modification of existing laws to enable AI innovation in an effective but still responsible way. The government has clearly stated it will rely on successful pilot programmes not only to identify which regulations need updating, but also to justify granting streamlined authority for making permanent changes , as opposed to following slower traditional legislative processes.

By adopting this regulatory strategy, the Government is seeking to test the regulators themselves, so that if their interpretation of the law is too cumbersome, this can be ironed out through subsequent legislation. While it would be wrong to interpret this as aggressive deregulation, the intention is to identify and correct unnecessary regulatory barriers. Ultimately, the proposal seeks to create an AI-friendly regulatory landscape,  supportive of both innovation and responsible AI development.

A third way

The current direction of travel is very telling, and not only in terms of its distinctive approach, but also in terms of what it is not. Around the world, AI regulation has so far followed one of two competing approaches: comprehensive risk-based frameworks that seek to impose prescriptive requirements on providers and deployers of AI, or a hands-off stance that deliberately avoids any kind of guardrails that could create friction. 

The UK is taking an altogether different path in this regard. Despite some efforts to pursue the former via a Private Members’ Bill, and more recently through calls by UK parliamentarians to introduce regulation applicable to the most powerful AI systems, the Government has steered clear of any legislative intervention. However, it would be unfair to say that concerns about the potential risks of uncontrolled AI development and implementation are simply being ignored or dismissed.

What the UK Government seems to have chosen to do is observe how existing laws deal with AI’s potential risks while it mulls over the need to intervene more explicitly. Therefore, specific AI regulation cannot be ruled out, but such regulation may, on the one hand, seek to address risks to safety or other aspects of life and, on the other, eliminate or modify existing rules that are deemed unhelpful to the wider implementation of AI technology. Considering how quickly AI technology is developing, it is reasonable to assume that some form of UK AI-specific regulation may emerge in 2026.

Ultimately, heavy-handed, prescriptive AI regulation is extremely unlikely to emerge in the UK anytime soon, but as this influential technology evolves, rapid changes to the legal landscape are to be expected.