The House of Lords is asking the question again: does the UK need a cross-sector AI law? Lord Holmes of Richmond scheduled a short debate on June 4 to push His Majesty’s Government on whether a dedicated AI regulation bill makes sense. The answer from government has consistently been: no, not yet. The debate is whether “not yet” is the right answer, or whether it’s a comfortable dodge.

The UK’s position is well-documented. Rather than follow the EU’s horizontal AI Act — one law covering every sector — the government has asked existing regulators to develop sector-specific guidance under their existing powers. The FCA handles AI in finance. The ICO handles AI and data protection. The Care Quality Commission handles AI in health. The Competition and Markets Authority watches for AI-related competition harms. This is the “adaptive, proportionate, and forward-looking” approach the government describes. It’s also a approach that critics say leaves significant gaps.

The AI Opportunities Action Plan: Growth First

The substantive policy backdrop is the AI Opportunities Action Plan, published in January 2025 and accepted in full by Prime Minister Starmer’s government. Its centrepiece recommendations include AI Growth Zones — with the first site at Culham, Oxfordshire — designed to accelerate planning and grid connections for AI data centres. The government is targeting a twenty-fold expansion of public AI compute by 2030. A National Data Library is in development to unlock public sector data for AI training.

These are growth-oriented interventions, and they matter. The UK’s AI sector has genuine strengths — DeepMind, Synthesia, Auturi, and a strong research base — and the government is clearly trying to create an environment where those companies can scale without the compliance overhead that European competitors face under the AI Act.

The Regulating for Growth Bill: Sandboxes and Standards

King Charles III’s May 2026 King’s Speech included the Regulating for Growth Bill, which takes direct aim at regulatory friction. The Bill — part of 37 measures announced in the speech — creates regulatory sandboxes that allow businesses to test AI products in controlled conditions without full compliance with existing rules. The idea is to de-risk innovation by giving companies a space to learn what works before committing to full-scale deployment.

This is a genuinely interesting approach. Regulatory sandboxes have worked well in fintech — the FCA’s sandbox produced real innovations and real learnings. Applying the model to AI is sensible in principle. The question is whether AI sandboxes can replicate that success, given the speed of the technology and the genuinely different risk profiles involved.

What’s notably absent from the King’s Speech: any mention of further AI legislation beyond the sandbox experiment. The cross-sector AI regulation debate is happening, but the government’s answer is, in effect: we’ll deal with it sector by sector, and meanwhile let’s build more data centres.

Why the Sector-Based Approach Has Real Limits

The sector-based model has a structural problem that the Lords debate was probing: what happens when AI harms don’t fit neatly into a single regulator’s jurisdiction? A hiring algorithm that discriminates in employment (FCA-adjacent?) while also violating data protection rules (ICO) while also potentially running afoul of consumer protection law (which regulator?) — who coordinates? Who has authority?

The AI Opportunities Action Plan acknowledged this gap and recommended a Central AI Risk Function within the Department for Science, Innovation and Technology to provide cross-sector coordination. Whether that function has real teeth or becomes a talking shop is the critical question. The Lords debate suggests Parliament is watching closely.

House of Lords short debate on cross-sector AI regulation: June 4, 2026. King’s Speech: May 13, 2026.