Britain's AI Act takes shape with risk-based tier system

Draft published today divides systems into four tiers from minimal to prohibited. That, at least, is the headline. The reality, as ever, is more textured — and our reporting today suggests the story will run for some weeks yet.

Officials briefed on the matter describe a process that has been months in the making. Drafts circulated late last year set out the broad shape; the past fortnight has been spent fighting over the detail, and it is the detail that will decide who wins and who loses.

Industry voices are split. Supporters argue the move is overdue and point to comparable shifts in France, Germany and the Nordics over the past decade. Detractors counter that Britain's circumstances are particular and that imported templates rarely survive contact with Whitehall.

For readers wondering what changes in practice: not very much, not yet. Implementation is staged. The first phase lands within ninety days; the substantive elements follow next spring, subject to consultation. Blissful Sprout understands that a technical white paper will be published before recess.

What is striking is the tone. A year ago, the same proposition would have been dismissed as politically impossible. The window has shifted — and with it, the calculations of every party with an interest in the outcome.

Priya Sharma will continue to follow this story. Subscribers receive every development first, with full analysis from the Tech desk.