Skip to main content
Back to Research

Tesseract Foundational Research: AI Governance

The closest thing to a public register of government AI, as a corpus

The UK now asks its public bodies to publish how they use algorithmic and AI tools in decisions that affect people. The result, the Algorithmic Transparency Recording Standard, is the nearest thing to a public register of government AI. It is published one record at a time, which makes the individual disclosures visible but the pattern invisible. This assembles the whole set into one corpus.

136
ATRS records
73
distinct public bodies
DSIT, DWP
most disclosures
OGL v3.0
open licence

The direction of travel

Transparency about public-sector AI has moved from aspiration to expectation. The Algorithmic Transparency Recording Standard, developed by DSIT and the Central Digital and Data Office, is now the mechanism by which departments and arm’s-length bodies disclose the algorithmic tools they use, and its use is being extended across government. Alongside the frameworks that govern this space, the EU AI Act, ISO 42001, the NIST AI RMF, the ATRS is where UK practice becomes visible. Reading it as a whole is how you see where government AI actually is.

The method: assemble the register

We harvested every published ATRS record from GOV.UK and structured each one: the publishing body, the tool, its description and its date. What was 136 separate pages becomes a single dataset spanning 73 distinct public bodies, led by DSIT, DWP, the Money and Pensions Service, DfE, DESNZ and DBT. The corpus makes answerable the questions no single record can: which parts of government are most transparent, which kinds of tools recur, and how disclosure is spreading.

DSIT
6
DWP
6
Money and Pensions Service
5
DfE
5
DESNZ
5
DBT
5

Public bodies with the most ATRS disclosures.

Where this goes

Captured over time, the corpus becomes a live index of government AI adoption and transparency, the base for a disclosure-growth measure and for classifying tools by risk and function. It is the public-register companion to Tesseract’s AI-governance work and to our open AI-governance tooling: transparency is only useful when someone reads the whole of it.

Independent, self-initiated open research. Contains public sector information licensed under the Open Government Licence v3.0.

Explore the corpus

All 136 ATRS records structured, and the reproducible harvester.