UK proposes AI rules through existing regulators

The UK government proposed a framework for AI regulation on 29 March, putting existing regulators in charge of applying five shared principles within their sectors.

Photo: 117PXL, Union flag in London (2022). CC BY-SA 4.0. Cropped/resized; this adaptation is also CC BY-SA 4.0.

The principles cover safety, transparency, fairness, accountability and routes to challenge harmful decisions. Regulators would tailor their approach to how AI is used.

The white paper makes the enforcement boundary clearer. Principles would initially be non-statutory, with regulators relying on existing powers. A later duty would require regulators to consider the principles while retaining discretion. The government could drop that duty if the initial approach works.

Annex A shows what implementation could involve: information about training data and decisions, regular model reviews and clearer complaint routes. These are proposed considerations for regulators, whose guidance would explain how the principles fit existing sector rules.

The initial sandbox pilot would cover one sector with multiple regulators. Its preferred service is customised advice; a safe testing environment would be explored later. For developers crossing regulatory boundaries, the proposed offer is help navigating existing rules, while regulators’ legal reach stays unchanged.

The government announced £2 million for sandbox support. The consultation is open until 21 June.