Inside the King's Speech AI Plan: What the Regulating for Growth Bill Means for UK Workflow Automation

37 — Bills in the 13 May 2026 King's Speech, zero standalone AI Act  ·  £500m — Sovereign AI Fund confirmed by Liz Kendall at RUSI, 28 April 2026  ·  HC 684 — Treasury Select Committee report accusing BoE, FCA and HMT of 'wait-and-see' on AI  ·  29 May — ICO consultation close on Article 22 automated decision-making reform

On 13 May 2026 the King opened Parliament with 37 bills, a copper-bottomed Police Reform Bill, an Employment Rights tidy-up, a Planning and Infrastructure Bill that finally tries to staple data centres to the grid, and one conspicuous absence: a horizontal AI Act. Anyone expecting Labour to ape Brussels' AI Act, or to translate the 2024 manifesto line about 'binding regulation' into a single tablet of stone, was instead handed the Regulating for Growth Bill, a statutory framework for cross-sector AI sandboxes that the Department for Science, Innovation and Technology (DSIT) is now calling the 'AI Growth Lab'. The Tech Secretary Liz Kendall, fresh from her RUSI keynote on 28 April, told the chamber the government would 'thread AI through sector reform rather than freeze it in horizontal law'. AI Minister Kanishka Narayan rolled the same line out across every London think tank that would have him.

The decision is more consequential than the absence of an AI Act suggests. For any London CFO drawing up a 2026-27 workflow automation budget, the regulatory mosaic Labour has chosen, statutory sandboxes plus sector regulators plus an ICO consultation closing on 29 May plus a Critical Third Party (CTP) fight between the Treasury Select Committee and HMT plus a CMA Strategic Market Status (SMS) probe into Microsoft, is now the operating environment. It is the environment in which procurement, build-versus-buy and US-cloud concentration decisions will be litigated for the rest of this Parliament. As an AI Agency London with deployment exposure to all of the above, we will, with our standard editorial cough, declare an interest: BraivIQ sells Workflow Automation London engagements that increasingly depend on which sandbox a client can credibly enter, and on how the CTP regime lands on the hyperscalers we deploy on top of.

What follows is the definitive read of the King's Speech AI plan as it actually affects mid-market and enterprise buyers in the United Kingdom. We cover the Regulating for Growth Bill mechanics, Lord Holmes's Henry VIII attack, the Treasury Select Committee's brawl with the regulators, the CMA's Microsoft SMS investigation, Kendall's £500m Sovereign AI Fund and its named recipients, the ICO's late-May guidance pile-up, and the Lib Dem flank attack on the $200bn US Tech Prosperity Deal. We close with a 90-day action list for boards that need to convert this mosaic into a procurement decision before the summer recess.

Why the Regulating for Growth Bill Matters More Than an AI Act Would Have

A statutory sandbox is, in plain English, a permission slip with a deadline. The Regulating for Growth Bill, as drafted, allows the Secretary of State to designate a use case, a sector and a participating regulator (FCA, MHRA, Ofcom, Ofgem, ICO, CAA, ORR, CMA), and to disapply or modify specified statutory provisions for a fixed term, typically twelve to twenty-four months, in return for monitoring, reporting and an automatic review clause. The model is descended from the FCA Regulatory Sandbox launched in 2016 and the MHRA AI Airlock pilot of 2024, but goes further: for the first time, the disapplications can cut across primary legislation, which is the constitutional point on which Lord Holmes has hung his case.

For a London CFO, the operational reading is straightforward. If your workflow automation programme touches financial services advice, clinical triage, transport scheduling, energy demand response or any regulated decision, your route to scale in 2026-27 now runs through whether your use case is sandbox-eligible. That is a different procurement question to 'is the model accurate'. It is a question about whether your AI Agency London partner can write a credible sandbox application, prove monitoring controls and absorb the reporting burden. We are already seeing tender requirements at large UK insurers and NHS trusts beginning to ask for it.

The Political Mosaic: Five Fights That Will Decide UK AI Procurement

1. Lord Holmes vs the AI Growth Lab

Lord Holmes of Richmond used the 26 March 2026 Lords debate on AI regulation to argue that the sandbox-by-designation model is a Henry VIII clause in disguise, ministers using delegated powers to rewrite primary legislation. He has reinforced the attack post-King's-Speech, joined by Baroness Stowell on constitutional grounds and Lord Clement-Jones on rights grounds. The Lords Library briefing of 9 May 2026 sets out the constitutional argument in detail. The practical implication for buyers is that the Bill is likely to be amended in the Lords with sunset clauses, parliamentary affirmative procedure for each designation, and rights-impact assessments, all of which will slow but not stop the sandbox process.

2. Treasury Select Committee vs HMT, BoE and FCA

The Treasury Select Committee, chaired by Dame Meg Hillier, published report HC 684 on 20 January 2026 accusing HMT, the Bank of England and FCA of a 'wait-and-see' posture on AI in financial services. The 18 March HMT response and 1 April BoE/FCA response declined to commit to designating AWS, Microsoft Azure or Google Cloud as Critical Third Parties for AI-specific services, despite the CTP regime under the Financial Services and Markets Act 2023 being designed precisely for this concentration risk. The 27 May 2026 joint statement from BoE, HMT and FCA warning on frontier AI cyber-exploit speed reads, between the lines, like an admission that the wait-and-see posture has run out of road. UK financial services CTOs should assume CTP designation for AI services is now a question of when, not whether.

3. CMA vs Microsoft (and the AWS reprieve)

The Competition and Markets Authority opened a Strategic Market Status investigation into Microsoft in May 2026 covering cloud, productivity AI and the Azure-OpenAI stack. Notably, AWS escaped this round, in part because the CMA's December 2025 cloud market study found AWS's egress and committed-spend practices materially less restrictive than initially feared. The DRCF (the joint CMA, ICO, Ofcom, FCA forum) published 'The Future of Agentic AI' on 17 April 2026, warning that agents acting on behalf of users may not be 'faithful servants' where principal-agent incentives diverge. For Workflow Automation London projects depending on Microsoft Copilot, M365 and Azure OpenAI, this is the single most important regulatory development of 2026: behavioural and structural remedies are on the table.

4. The Sovereign AI Fund and the Compute Question

Liz Kendall used her RUSI speech on 28 April 2026 to confirm the UK AI Hardware Plan and a £500m Sovereign AI Fund. Confirmed equity-and-grant recipients include Callosum (interpretability), Ineffable Intelligence (frontier reasoning), and Isomorphic Labs (drug discovery), with subsidised compute access for Cosine, Doubleword, Odyssey, Prima Mente, Cursive and Twig Bio. Datacenterdynamics.com has detailed the compute allocations across Bristol, Edinburgh and Culham. The fund is small compared to the $200bn US Tech Prosperity Deal but signals that Labour is hedging horizontal sovereignty rather than competing dollar-for-dollar.

5. The ICO's Late-May Pile-Up

In May 2026 the Information Commissioner's Office published FOI guidance on public sector AI procurement transparency, a five-step AI cyber resilience plan, and ran a consultation on reform of Article 22 (automated decision-making) under the Data (Use and Access) Act 2025, closing 29 May. The consultation outcome will determine whether 'meaningful human review' in workflow automation contexts means a human in the loop, a human on the loop, or a human merely on call. The answer changes how every customer service, HR screening and credit decision automation we build at BraivIQ needs to be designed.

What This Means for Build-Versus-Buy and US Vendor Concentration

The central strategic question Labour's mosaic poses to a UK board is not 'should I use AI' but 'how concentrated is my AI supply chain on three US hyperscalers and one US model provider, and what is my exposure if any one of them is designated CTP, hit by SMS remedies, or constrained by Article 22 reform?' The Twobirds analysis of the Regulating for Growth Bill notes that sandbox designation is likely to favour use cases with clear UK data residency, identifiable model provenance and auditable decision logs, all of which are easier to demonstrate on a mixed estate with at least one UK-or-EU sovereign component than on a pure Azure-OpenAI stack.

This is not a call to rip out Microsoft. It is a call to build optionality. The Workflow Automation London engagements we are scoping for Q3 2026 increasingly include a sovereign-fallback design: a primary path on Azure-OpenAI or AWS-Bedrock, a secondary path on UK-hosted open-weight models (Mistral, Llama, the Cohere Command stack on Oracle UK South), and an orchestration layer that can route between them based on data classification and regulatory posture. That is the architecture a credible AI Agency UK should be offering, and increasingly the one regulators are quietly preferring.

The Liberal Democrat Flank Attack

The Liberal Democrats, through Daisy Cooper and Lord Clement-Jones, have rebranded their AI position as 'AI Digital Independence', explicitly framed against the $200bn US Tech Prosperity Deal signed by the Starmer government in September 2025. The party's amendment programme for the Regulating for Growth Bill will push for mandatory UK-residency triggers, mandatory open-weight options in public procurement above £5m, and a statutory right-to-explanation that goes further than ICO Article 22 reform. None of this will pass in its filed form, but it shifts the median amendment Labour will accept. UK boards procuring AI services for public-sector clients should expect open-weight clauses in tenders by Q4 2026.

The 90-Day UK Board Playbook

1. Map your Critical Third Party exposure. List every AI workflow that depends on AWS, Microsoft Azure or Google Cloud and rate it for substitution feasibility within twelve months. This is the document your audit committee will ask for by year-end.
2. Log every sandbox-eligible use case. Walk the Regulating for Growth Bill draft against your roadmap and identify candidates for FCA, MHRA, ICO or DRCF sandbox entry. Sandbox-eligible use cases will move twice as fast to production over the next two years.
3. Prepare for the ICO ADM final guidance. Audit every automated decision in your business against the consultation document closing 29 May 2026 and stand up a human review log before the final guidance lands in Q3.
4. Audit your US-cloud dependency and price the sovereign fallback. Quote a parallel architecture in which 20 to 30 per cent of inference can be served from UK-hosted open-weight models within nine months. Even if you never switch, the price of optionality is now a board-level number.
5. Recruit or retain a regulator-literate AI Agency London partner. The model layer is commoditising, the compliance layer is not. Pick a partner who can write a sandbox application, draft a CTP impact assessment and read an ICO consultation, because over the next eighteen months that is where margin and risk both live.

Sources

1. UK Parliament, King's Speech 2026 background briefing, 13 May 2026, parliament.uk
2. House of Lords Library, Briefing on AI Regulation and the AI Growth Lab, 9 May 2026, lordslibrary.parliament.uk
3. Treasury Select Committee, Report HC 684 'AI in Financial Services', 20 January 2026, with HMT response 18 March 2026 and joint BoE/FCA response 1 April 2026, parliament.uk
4. HM Treasury, Bank of England and FCA, Joint statement on frontier AI cyber-exploit risks, 27 May 2026, gov.uk
5. Department for Science, Innovation and Technology, Liz Kendall RUSI speech on the UK AI Hardware Plan and Sovereign AI Fund, 28 April 2026, gov.uk
6. Bird and Bird (Twobirds) analysis, 'The Regulating for Growth Bill and the AI Growth Lab', May 2026, twobirds.com
7. Competition and Markets Authority, Strategic Market Status investigation into Microsoft cloud and AI services, May 2026, gov.uk/cma
8. Digital Regulation Cooperation Forum, 'The Future of Agentic AI', 17 April 2026, drcf.org.uk
9. Information Commissioner's Office, FOI guidance on public sector AI procurement, AI cyber five-step plan, and Article 22 consultation closing 29 May 2026, ico.org.uk
10. The Actuary magazine, coverage of HC 684 and AI concentration risk in UK insurance, February 2026, theactuary.com
11. Data Center Dynamics, 'Inside the UK Sovereign AI Fund recipients', May 2026, datacenterdynamics.com
12. Financial Times, coverage of the CMA Microsoft SMS investigation and AWS reprieve, May 2026, ft.com
13. BBC News, coverage of Lord Holmes Henry VIII clause attack on the AI Growth Lab, March and May 2026, bbc.co.uk
14. Computer Weekly, analysis of the Liberal Democrat 'AI Digital Independence' position and the $200bn US Tech Prosperity Deal, April-May 2026, computerweekly.com