Estonia's Kratid programme runs a public, voluntarily-submitted register listing around 120 AI systems in use across roughly 60 government bodies, creating a searchable public inventory of state AI adoption — though participation remains optional and unaudited.
~120
AI implementations listed
~60
Public authorities contributing
Details
Maturity
Established
Promoter
Ministry of Justice and Digital Affairs / State Information Systems Authority (RIA) — Kratid programme
Period
2017–present
Region (NUTS)
EE00
Keywords
digital government, AI transparency, public administration
Context
Estonia's Ministry of Justice and Digital Affairs and the State Information Systems Authority (RIA) jointly run Kratid, a public-facing programme for coordinating artificial intelligence across the Estonian public sector, centred on a searchable online register (kratid.ee).
Activities
Government bodies voluntarily describe the AI systems they have built or deployed -- chatbots, computer-vision tools, forecasting models -- alongside purpose, sector and deployment status.
Results
As of the latest published listing, the register documents roughly 120 AI implementations contributed by close to 60 public authorities, spanning projects launched between 2017 and 2022 across economy, security, environment, education, law, health and cultural-sector agencies.
Conclusions
Submission is voluntary rather than legally mandated, so the register almost certainly understates the true number of AI systems in government; Estonia has not published an independent audit to verify completeness, and it lacks the structured risk-classification fields found in more rigorous registers such as Brazil's judicial Sinapses catalogue.
Implementation
Indicative cost
Low (< €50k) — Government-run public register; running costs not disclosed.
Time to results
Long (> 3 years) — Covers projects launched 2017-2022, register maintained on an ongoing basis.
Staffing & skills
Ministry of Justice and Digital Affairs (Estonia), State Information Systems Authority (RIA)
Conditions for success
single searchable public register lowering the barrier for agencies to disclose AI use
Common failure modes
voluntary submission likely understates true AI adoption
no independent completeness audit
lacks structured risk classification or impact-assessment fields