Coordinated by Israel's State Comptroller, 12 EUROSAI audit institutions applied a shared 9-topic, 92-question framework to benchmark government AI readiness, finding ownership …
The UK Foreign Office uses Microsoft AI Builder and GPT-4.1 to auto-classify consular correspondence, publishing UAT accuracy of 73-95% by task. Officials …
Nepal Customs built an in-house machine-learning module inside ASYCUDA World, now running in the six largest customs offices, screening ~90% of commercial …
In 2024 Brazil's Piauí state made AI compulsory from 9th grade through upper-secondary -- the Americas' first such curriculum -- reaching 90,000+ …
DSIT and DCMS commissioned the Behavioural Insights Team to run matched AI-assisted and human-only rapid evidence reviews on the same brief. The …
Brazil's National Treasury replaced a 1,000-hour manual process with a machine-learning classifier that codes subnational spending to the UN's COFOG standard in …
Bangladesh's a2i, GiveDirectly, UNDP and UC Berkeley piloted ML poverty-targeting from call-detail records across 106,200 Cox's Bazar households, but a World Bank …
Mexico's statistics institute INEGI is scaling BERT/FastText-based AI coding from its household income survey to the national employment survey, formally targeting up …
Zambia's national statistics agency and the IGC tested seven LLMs against 1,059 human-coded Labour Force Survey responses: the best models beat enumerators …
England's Department for Education and the Incubator for AI selected up to 8 edtech/AI firms, backed by £300,000 each, to co-design AI …
Great Britain's grid operator NESO deployed Quartz Solar, an AI forecasting tool built with non-profit Open Climate Fix. It halved solar-forecast errors, …
Turkey's Revenue Administration used a GIS-based, AI-assisted value-matching system to scan over 16,000 properties in 2025, flag 9,150 owners for unexplained value …
An open-source Analyzer module, built by STACC for Estonia's Information System Authority (RIA), applies adaptive statistical models to X-Road's data-exchange logs to …
WFP's HungerMap LIVE uses machine-learning "nowcasts" to estimate food insecurity where ground surveys are impossible. In Sudan's civil war, it flagged 21.2 …
Germany's Rhineland-Palatinate bought every teacher in the state a free fobizz licence for AI-assisted lesson planning and content creation; by end-2024 nearly …
Since 2021, Los Angeles County's Homelessness Prevention Unit has used a California Policy Lab model built on 580 administrative-data points to rank …
A World Bank team built 110m-resolution poverty maps for São Tomé and Príncipe by combining satellite imagery with machine learning, giving planners …
Iraq's Customs Authority deployed UNCTAD's ASYCUDA World, whose newest generation adds AI-based risk-scoring to sort shipments into inspection channels. Revenue roughly doubled …
Politecnico di Milano's Perivallon project uses neural networks on satellite imagery to flag illegal waste dumps with over 90% accuracy, piloted across …
Greater Cambridge Shared Planning and the University of Liverpool built PlanAI, a large language model trained on 15 years of planning consultations, …
A World Bank-designed randomized controlled trial gave about 800 Edo State, Nigeria senior-secondary students six weeks of after-school, teacher-guided ChatGPT tutoring in …
Palau's Bureau of Customs and Border Protection deployed UNCTAD's ASYCUDAWorld 4.4 with an AI "dynamic selectivity" risk engine in January 2024. Within …
Brazil's PNLD textbook programme tested a CNN sorting book images as sharp, defocus-blurred or motion-blurred in a two-arm 1:1 randomised study with …
Under OMB Memorandum M-24-10, US federal agencies must publish a machine-readable annual inventory of every AI use case. The 2025 edition logged …
Open full copilot Grounded in cited practices — always check the sources.
