Summary
- Spotlight on Corruption calculated that 84% of ministerial data-centre stakeholder meetings were with businesses or trade associations.
- The organisation reports no ministerial or senior official meetings with environmental NGOs in its defined sample.
- The findings are the campaign group’s analysis and should not be treated as an official audit of all consultation activity.
Spotlight on Corruption found that businesses and private-sector trade associations accounted for 84% of ministerial stakeholder meetings on data centres between July 2024 and June 2026. The organisation published its analysis of engagement over AI Growth Zone policy on 27 October, raising questions about the balance of access to senior UK decision-makers.
The 27 October report, titled Zoned Out, examines engagement with government during the formulation of policies designed to accelerate AI infrastructure development. Its findings are the campaigning organisation’s analysis of meeting information and related material, not an independently audited count of every discussion about digital infrastructure.
Spotlight on Corruption reports that ministers met only two community groups in the period examined, both after their areas had been designated as AI Growth Zones. Its dataset identified no environmental-NGO meetings involving ministers, permanent secretaries and directors general. These are findings about the recorded meetings within the study, not a complete census of every government contact.
Those numbers describe a specified group of officials and an identified period. They should not be converted into claims that no civil servant, planning officer or local authority held any community meeting anywhere in the UK, or that the investigation covers every channel through which public views may have reached government.
Policy engagement and infrastructure consent
AI Growth Zones are intended to concentrate infrastructure development and supporting investment in selected locations. The physical projects associated with them can require large electricity connections, water assessments, transport and construction works, telecommunications links and local planning decisions.
Government policy designation is not identical to granting planning permission for a particular data-centre building. Developers can still face site-specific processes, environmental assessments, grid applications and statutory consultation obligations. The investigation’s central concern is about engagement during national policy formation and selection of areas, rather than an assertion that every local consent has bypassed ordinary procedure.
Consultation conducted after an AI Growth Zone is designated may allow residents to comment on mitigation and project design without reopening the earlier decision about the area’s selection. The report argues that greater community involvement was needed before national policy priorities and locations were established.
The organisation also reports that ministers, permanent secretaries and directors general held only one meeting with local authorities before the designation of their areas as AI Growth Zones. That claim should be assessed in the same terms as the underlying dataset, since local authorities may have participated in technical discussions or other processes outside the meetings counted.
According to the report, the government’s Sustainability Working Group included no environmental organisations and met once between its creation in January 2025 and July 2026. The organisation also says national environmental and social impact assessments for the overall AI Growth Zone approach had not been released at the time of publication.
Power, water and public accountability
Electricity availability is a major practical constraint on data-centre delivery. An ambitious development zone can provide a planning or policy framework, but it cannot by itself energise a campus, provide transmission capacity or resolve the timing of reinforcement works.
Water supply and cooling demand also vary between projects. A facility using predominantly dry heat rejection has a different direct water profile from one relying substantially on evaporative cooling, while both require an assessment of local circumstances. General claims about the water intensity of AI computing should not substitute for project-specific demand forecasts and permits.
Construction and operational employment need similar care. Announced investment values are not equivalent to local economic benefit, and construction jobs are not the same as continuing operational posts. Identifying the likely local workforce, supply chain, tax base and infrastructure costs can help establish whether proposed community benefits are realistically supported.
Spotlight on Corruption recommends more transparent stakeholder-engagement plans, greater participation by public-interest groups and publication of relevant environmental and social assessments. It also calls for a pause in the programme until stronger criteria and safeguards are established. Those are recommendations from the investigating organisation, not current government policy.
The investigation separately identifies 38 ministers and senior officials who moved into roles at AI businesses, covering a different period from the meeting analysis. That figure does not establish wrongdoing by particular individuals or a breach of rules governing post-government employment.
The government’s response should be reported separately from the organisation’s criticism. Ministers have defended AI Growth Zones as a way to accelerate investment, jobs and computing infrastructure, while the investigation questions the balance and timing of access to decision-makers. Neither position determines the technical feasibility or planning legality of an individual development.
Spotlight on Corruption’s dataset covers specified senior-government meetings between July 2024 and June 2026; its separate employment-movement figure uses a different period. The findings raise questions about the national policy process and provide a basis for demanding greater disclosure, but do not establish that any particular data-centre consent is invalid.
