UK AI Infrastructure Governance: Commitments, Zones, and Accountability Gaps

Executive Summary

Between January 2025 and January 2026, sixteen parliamentary questions were tabled across both Houses concerning artificial intelligence infrastructure in the United Kingdom. Fourteen originated in the House of Commons, directed to the Department for Science, Innovation and Technology (DSIT); one Lords question addressed environmental and community dimensions of AI infrastructure development; and two questions — formally classified under overseas trade — touched on sovereign AI infrastructure in the context of UK-Canada relations. The dominant ministerial voice across this period was Feryal Clark (Commons), with Lord Vallance of Balham, Baroness Jones of Whitchurch, and Baroness Lloyd of Effra providing Lords responses. This brief assesses the substance, consistency, and accountability value of those responses.

Volume and Pattern Analysis

The concentration of questions is itself instructive. Of the sixteen items, twelve were asked between January and May 2025, reflecting the immediate parliamentary appetite generated by the publication of the AI Opportunities Action Plan on 13 January 2025. Three questions were tabled by the same MP, Kanishka Narayan, on a single date — 25 March 2025 — covering cross-government compute planning, energy and grid access, and the proximity of AI Growth Zones to research centres. This clustering suggests deliberate parliamentary scrutiny rather than incidental interest, and indicates that the answers provided to earlier questions were not considered sufficient by Members seeking granular operational detail.

Lord Taylor of Warwick accounts for three of the Lords questions, spanning February 2025 to January 2026, and demonstrates a sustained interest in site selection criteria, foreign investment attractiveness, and — most recently — environmental assessment and community engagement in light of legal challenges to major projects. The evolution of his questioning from process to legal risk signals a sharpening of parliamentary concern as the programme matures.

Ministerial Response Quality

Ministerial responses across this dataset are broadly consistent in tone and content — arguably too consistent. Several answers deploy near-identical formulations. The phrase “securing the UK’s position as a global leader in AI innovation” appears, with minor variation, in at least five separate responses. The framing of AI Growth Zones as delivering “substantial regional and national benefits, such as upskilling and employment opportunities” recurs across multiple answers to different MPs. While such consistency may reflect departmental briefing discipline, it also limits the analytical value of the responses as accountability instruments: Members asking distinct questions receive substantively similar answers.

Where responses do differentiate, they tend to reference process milestones rather than outcomes. The announcement of Culham — home to the UK Atomic Energy Authority — as the first designated AI Growth Zone is the most concrete factual disclosure across the entire dataset (UIN 51912, 13 May 2025). Beyond this, ministers confirm that over 200 expressions of interest were received following the February 2025 open call, and that a formal rolling application process has been opened. No further confirmed zone designations are referenced in any response within this dataset.

Key Commitments and Timelines

The parliamentary record identifies several specific commitments and timelines against which delivery can be assessed:

Spring 2025 — Formal selection process launch: Lord Vallance of Balham stated in February 2025 that “the Government will lay out the formal selection process in spring” (UIN HL4909). Feryal Clark confirmed in January 2025 that “the selection process will open in spring 2025” (UIN 24689). The May 2025 response confirming the formal process had opened (UIN 51912) suggests this commitment was broadly honoured, though no specific spring date is confirmed in the responses.

National Policy Statement on data infrastructure: The March 2025 response to Kanishka Narayan (UIN 41205) commits DSIT to “producing a National Policy Statement on data infrastructure to guide future planning decisions.” No subsequent question in this dataset confirms publication or a publication date.

Planning and Infrastructure Bill: Referenced in May 2025 (UIN 54307) as the legislative vehicle for streamlining planning approvals for AI infrastructure and enabling larger data centres to enter the Nationally Significant Infrastructure Projects consenting regime. The Bill’s parliamentary progress is not addressed in any subsequent response within this dataset.

AI Energy Council: Mentioned in the March 2025 response on grid access (UIN 41202) as the forum through which “bold clean energy solutions” — including next-generation renewables and small modular reactors — will be explored. No further detail on its membership, meeting cadence, or outputs appears elsewhere in the dataset.

UK-Canada sovereign AI infrastructure cooperation: Two September 2025 responses (UIN 77027 and UIN 76570) reference cooperation with Canada on “sovereign artificial intelligence infrastructure” as part of the UK-Canada Economic and Trade Working Group’s terms of reference. This represents the sole international dimension of AI infrastructure governance in this dataset, and no subsequent update on outcomes is provided.

Rhetoric Versus Reality

The central tension in this parliamentary record is the disjunction between the ambition of the Government’s stated AI infrastructure programme and the paucity of confirmed, operational outcomes within the timeframe covered. The AI Opportunities Action Plan, published in January 2025, is cited in virtually every ministerial response as the authoritative framework — yet the plan itself is a statement of intent rather than a delivery ledger. Parliamentary questions that probe implementation — on devolved nations (UIN 37150), on West Midlands engagement (UIN 32938), on research centre proximity (UIN 41200) — receive responses that describe process and aspiration without confirming tangible delivery.

The environmental and community engagement dimension, raised by Lord Taylor of Warwick in January 2026 (UIN HL13951), receives a response that points to existing planning frameworks as the safeguard, without acknowledging the specific legal challenges referenced in the question. This is a notable deflection: rather than addressing the adequacy of current frameworks in light of live legal contestation, the minister reasserts the frameworks’ existence. For an independent analytical observer, this pattern — process cited as evidence of progress, existing frameworks cited as evidence of adequacy — is characteristic of a governance posture more comfortable with narrative management than with substantive accountability.

The repeated invocation of the Mansion House Accord, the British Growth Partnership, the National Wealth Fund, and the “over £100 billion of additional capital” figure (UIN HL8132) in the context of AI infrastructure attractiveness is similarly notable. These are wider financial reform instruments, not AI-specific mechanisms, and their citation alongside AI infrastructure commitments risks conflating distinct policy streams in a manner that inflates the apparent policy density of the Government’s response.

Strategic Intelligence Assessment

This parliamentary dataset captures the UK Government’s AI infrastructure programme at an early implementation stage — past launch, not yet at verifiable delivery. The Culham designation as the first AI Growth Zone is a genuine milestone. The expression of interest process generating over 200 responses reflects real market and local authority appetite. The cross-departmental coordination described — between DSIT, MHCLG, DESNZ, and the National Energy System Operator — is structurally appropriate if operationally unverified.

However, the accountability deficit is clear. Commitments to a National Policy Statement on data infrastructure, a functioning AI Energy Council, confirmed additional Growth Zone designations, and progress on the Planning and Infrastructure Bill are either unconfirmed or absent from subsequent ministerial responses. The rolling application process for Growth Zones — while operationally flexible — also defers the moment of public accountability for selection decisions, making parliamentary scrutiny more difficult.

The entry of sovereign AI infrastructure into UK-Canada trade discussions (September 2025) is an analytically significant development that warrants closer monitoring: it suggests the Government is beginning to frame AI infrastructure as a geopolitical and trade asset, not merely a domestic industrial policy instrument. This framing, if sustained, will have implications for investment screening, national security review, and the coherence of the UK’s international AI governance positioning.

ISAR Global assesses that the UK Government’s AI infrastructure governance, as evidenced in this parliamentary record, is characterised by process momentum without proportionate outcome accountability. The rhetorical architecture is sophisticated and internally consistent. The delivery architecture remains, on parliamentary evidence, substantially unverified.