Open Data Beer #34 · LIIP Zurich - June 4, 2026

Public data reporting in the age of AI

Four cases where public-data portals, APIs, and registers became accountability journalism, and what changes when agents can monitor those signals every day.

Tom Vaillant Buried Signals
Before the cases

Public access is necessary. Reporting makes it matter.

Open data is the foundation layer: filings, permits, contracts, registers, court records, environmental measurements, and minutes — where public-interest leads first appear.

Open data
+
Cleaning
+
Context
+
Human verification
=
Accountability
Case 1 - France

Le Monde made groundwater contamination searchable.

Public data layerFrench groundwater measurements via public water data infrastructure; Le Monde republished the cleaned dataset on data.gouv.fr.
QuestionWhich contaminants are present in groundwater, and where?
MethodologyNormalize measurements, match contaminants to thresholds, aggregate by station, then map exceedances locally.
ImpactA national environmental story readers could inspect by place.
Case study Hub'Eau API Data source
data.gouv.fr dataset published by Le Monde for the groundwater investigation
data.gouv.fr reuse dataset published by Le Monde
Case 2 - Norway

Follow public welfare money into private providers.

Public data layerCompany registry APIs, public accounts, ownership records, and kindergarten-sector data.
QuestionHow did public funding become a commercial arena?
MethodologyJoin provider accounts, company registers, ownership, properties, subsidies, and sector data.
ImpactData-SKUP recognition for small-newsroom accountability work.
Case study Brønnøysund API Sector data
Utdanningsnytt article screenshot showing Data-SKUP award
Utdanningsnytt / E24 Data-SKUP coverage
Case 3 - UK / EU

A ferry company with no ferries appeared in procurement data.

Public data layerTED notices and open contracting data made emergency public spending visible across borders.
QuestionWho received emergency Brexit transport contracts?
MethodologySearch awards, isolate the anomalous supplier, then verify capability against public records.
ImpactAn anomalous award became a national procurement controversy.
Case study TED source Open contracting data
Open Contracting Partnership article screenshot about a controversial Brexit contract
Open Contracting Partnership case study
Case 4 - Switzerland

Roll-call votes refuted a Swiss party's "neither left nor right" slogan.

Public data layerOpenParlData.ch exposing every nominal vote of Geneva's Grand Conseil. Le Temps analyzed 692 scrutins from 27 January 2022 to 4 November 2025.
QuestionTwenty years in, where does the MCG actually sit on Geneva's political map?
MethodologyW-NOMINATE scaling on every roll-call vote since 2022 places each party and legislator on a left-right axis from voting similarity alone, beyond stated positions.
ImpactQuantitative evidence that MCG votes in lockstep with Le Centre and LJS — a synchronized center-right bloc, not the centrist outsider it claims.
Le Temps case study OpenParlData.ch W-NOMINATE method
Le Temps W-NOMINATE chart showing Geneva Grand Conseil parties positioned on a left-right axis, source OpenParlData.ch
Le Temps — W-NOMINATE positions of Geneva parties, source OpenParlData.ch

The same AI technologies that consolidate power can be turned around to hold it accountable.

One journalist with the right tools can now investigate at the scale of a traditional newsroom, if public data is accessible.

Public data reporting in the age of AI

Accessible data gives AI something accountable to work with.

1Collect

Portals, registers, agendas, filings, PDFs, APIs.

2Connect

Link entities, places, contracts, payments, and people.

3Question

Find anomalies, contradictions, missing context, and follow-ups.

4Verify

Preserve evidence, call sources, test assumptions, publish methods.

iMEdD article screenshot mentioning iTromso and DJINN
iMEdD notes on iTromso / DJINN and public-document leads
AI demo - iTromso / DJINN

DJINN turns municipal PDFs into leads reporters can verify.

Accessible data layerScrapers and APIs collect over 12,000 PDFs a month from municipal archives.
Old hospital building illegally converted into apartments Floating-sauna permit violations and police complaints
Model jobRank relevance, summarize filings, extract key information, and suggest issues.
MethodologyJournalists train the model with thumbs-up/down labels: is this a news story or not?
Human jobFollow the lead, verify the filing, call people, and report the story.
Media City Bergen iMEdD / Nordic AI in Media
About me

Tom Vaillant

JOURNALIST & TECHNOLOGIST

Buried Signals is an investigative and technology studio in partnership with Indicator. Using AI to accelerate investigations, disclose methodologies, and build open tools for journalists and researchers.

CUNY AI Journalism Lab Builders Cohort Pulitzer Center grantee Collaborations with Indicator, Bellingcat, Big Local News and MuckRock
buriedsignals.com/about tom@buriedsignals.com
Tom Vaillant
The journalism proves the tools. The tools make the journalism repeatable.
The ecosystem

Four tools. Built, tested, used.

Buried Signals builds practical investigation infrastructure with open methods and partner newsrooms.

Spotlight

Agentic investigations: research, verification, evidence, and drafting workflows.
Scoutpost

AI scouts for public pages, social posts, civic documents, and recurring beats.
OSINT Navigator

An AI guide to investigation tools, methods, and source selection.
Mycroft

Research support that turns weak signals into investigation leads.

What I am building with Indicator

Data Navigator turns open portals into queryable reporting tools.

A query-first catalogue for public data sources: ask in plain language, choose a source, get records back with source URLs preserved.

Indicator Data Navigator repo
Data Navigator Query public data sources
Find Norwegian companies named Equinor
Brønnøysund companies USAspending awards GLEIF LEI records
Recordname, jurisdiction, company number, source URL
ExportJSON and CSV-ready results for reporting workflows
Buried Signals
In partnership with Indicator

Investigations with AI. The tools to run your own.

buriedsignals.com LinkedIn