OpenPolice began as the UK's first independent, community-driven platform for police accountability — a way for the public to access information about police forces, courts and conduct matters, built on open data and citizen participation. The same team had built police-accountability platforms in the United States and other countries, and the ambition was to bring that model to the UK.

It was never a vigilante or anti-police project. The goal was constructive transparency: giving the public a clear view of policing, and supporting the principle of policing by consent.

In the US and elsewhere, Freedom of Information regimes release officer rosters, disciplinary outcomes and related records in bulk and in machine-readable form. Those platforms could launch fully populated — the dataset existed on day one, and the work was to organise, verify and present it.

The UK's Freedom of Information Act 2000 works differently. Forces are not required to release officer-level rosters or individual conduct data in bulk, and such requests are routinely refused under personal-data exemptions. The foundational dataset the platform was architected around simply never existed for the UK.

To work around the missing FOI pipeline, the UK platform was built around crowdsourcing: citizens submitted records and a tiered community of moderators verified and curated them. The response was genuinely overwhelming, and we're grateful to everyone who took part.

But a hand-assembled dataset has hard limits. It couldn't match the completeness of records released in bulk under law, every submission needed human review to stay accurate and lawful, and the moderation workload grew with the contributor base. Over time, the effort to keep the data trustworthy outpaced what a volunteer community could sustain.

Compensating for sparse, crowdsourced data made the platform unusually compute- and memory-hungry — entity resolution and deduplication, image and document processing, large in-memory search indexes, and AI inference for classification and moderation.

These are RAM-bound workloads, and the sustained rise in memory prices pushed both hosting and inference costs up sharply. With no bulk FOI data to offset the per-record cost, every additional record made the UK platform more expensive to run — the opposite of how the US model behaved. For a free, community-funded public service, that cost curve wasn't sustainable.