HyperscalerWatch — Editorial Principles and Public Trust
Version: `public-trust-v1` · Effective 18 November 2026.
HyperscalerWatch is an independent, free public information resource about data centres and their development across the United Kingdom, including Northern Ireland. Our purpose is to help people understand data-centre infrastructure and the different considerations surrounding its development.
HyperscalerWatch must never advocate for or against data centres, individual developments, operators, planning applicants or community groups. Our objective is greater public understanding, not public persuasion.
1. Inform, don't persuade
Present evidence and explain its significance without telling readers what conclusion to reach. Do not adopt promotional industry messaging or opposition campaigning language.
2. Be curious about both benefits and concerns
Explain documented economic and infrastructure benefits alongside environmental, planning and community considerations. Do not manufacture false balance. Give claims prominence according to their relevance and supporting evidence, not an arbitrary requirement for equal positive and negative coverage.
3. Distinguish facts, forecasts and opinions
Clearly differentiate established facts from official records; developer statements and projections; independent technical findings; planning authority assessments and decisions; community objections and support; unverified claims and unresolved questions. Never turn a prediction into an observed outcome or an allegation into a finding. Identify the author, document date, observation date and phase where relevant. Publication by an authority does not make every statement in its report an authority finding.
4. Make technical information accessible
Write in clear, natural British English for intelligent non-specialist readers. Explain unfamiliar terms without patronising readers. Use meaningful comparisons only when valid, with assumptions and limitations disclosed. Avoid sensationalism, jargon, corporate marketing language and generic AI-generated prose. Proposed IT power, maximum site demand, connection voltage and measured electricity consumption remain different quantities; drainage is not cooling-water consumption and peak staff is not total employment.
5. Be transparent about uncertainty
If water use, the eventual operator, job numbers or another important detail isn't publicly established, say so clearly. Distinguish information that has not yet been researched from information that research has not found. Qualify a negative finding by the documents and date checked. Failed access is an access gap, not proof of absence. Don't use missing information to imply wrongdoing or no impact.
6. Preserve the original evidence
Every material factual claim must be traceable to an appropriate source. Prefer original planning records, government information, technical documents and official statements. Identify who made a claim and when. Present relevant contradictions and changes over time without hiding them. Preserve claim-level citations, locators, source/retrieval dates, original units, reuse terms, scope and verification state. Restricted full text remains private; public pages use permitted neutral factual summaries and original citations.
7. Represent community perspectives fairly
Summarise both objections and supporting representations where documented. Describe the actual issues and evidence, rather than labelling people as anti-development or pro-industry. Do not publish unnecessary personal details of private individuals who submit planning representations. Representations are views or allegations unless a separate assessment establishes a finding. Do not infer how many people hold a view from how many research observations mention it.
8. Explain the planning process
Distinguish announced, submitted, approved, under construction and operational developments. Explain what planning approval does and does not establish. Show how applications, amendments and decisions relate to each project. A decision on an amendment is not proof of construction or operation. Keep separate campus phases and historical statuses identifiable.
9. Remain independent
Do not give any operator, developer or advocacy group favourable treatment. Apply the same research, verification and editorial standards to every project. Clearly disclose any future funding arrangements or potential conflicts of interest. Do not invent a funding declaration when no arrangement has been established; source attribution is not endorsement.
10. Correct mistakes transparently
Maintain dated research updates, sources and correction history. Automatically correct supported factual errors under the established verification process. Make it easy for readers to submit corrections or identify missing evidence. A reader report is a lead, not established evidence; verify it against appropriate sources. Preserve prior versions, correction/retraction dates and provenance while protecting private source material and personal details.
Application in the autonomous system
The same versioned rules apply to deterministic extraction, optional local-model extraction and public editorial summaries. Mechanical checks catch common advocacy, personal-data and assertion errors; they do not constitute a complete semantic understanding of every source. Failed claims are recorded privately for autonomous re-research, while qualifying claims continue to publish automatically. There is no mandatory human approval queue.
The public presentation keeps source IDs and makes attribution explicit. Relevant documented support and objections can both appear without inventing a counterclaim or imposing equal prominence. Research-category records distinguish findings, bounded searches with no finding, blocked access and work not yet undertaken. Corrections and supported updates follow existing versioning and verification rather than a separate editorial veto.