Search
187 posts across four languages. This searches titles, descriptions, topics, places and language — not the body text of the articles. Results are ordered by how many of your words match, then by where they matched (title, then description, then topic, place or language), then by date. Posts matching only some of your words are shown too, lower down and labelled.
2026
Assorted links for 29 August 2026
Nine links: which of England's two planning AI tools is actually live, what MIT's committee found had changed on campus in under three years, an Ebola nowcast in the Democratic Republic of the Congo, and the first evaluation of a proprietary model by someone who never saw its weights.
developmentevalsservice-deliveryUnited KingdomUnited StatesCDNigeriaIndiaSingapore
Enlaces varios para el 29 de agosto de 2026
Nueve enlaces: cuál de las dos herramientas de IA del sistema urbanístico inglés está realmente en marcha, qué constató la comisión del MIT en el campus en menos de tres años, una previsión de brote de Ébola en la República Democrática del Congo, y la primera evaluación de un modelo propietario por alguien que nunca vio sus pesos.
developmentevalsservice-deliveryUnited KingdomUnited StatesCDNigeriaIndiaSingapore
Liens divers pour le 29 août 2026
Neuf liens : lequel des deux outils d'IA du système d'urbanisme anglais fonctionne vraiment, ce que la commission du MIT a constaté sur le campus en moins de trois ans, une prévision d'épidémie d'Ebola en République démocratique du Congo, et la première évaluation d'un modèle propriétaire par quelqu'un qui n'en a jamais vu les poids.
developmentevalsservice-deliveryUnited KingdomUnited StatesCDNigeriaIndiaSingapore
Links diversos para 29 de agosto de 2026
Nove links: qual das duas ferramentas de IA do sistema de urbanismo inglês está de fato no ar, o que a comissão do MIT constatou no campus em menos de três anos, uma previsão de surto de Ebola na República Democrática do Congo, e a primeira avaliação de um modelo proprietário por quem nunca viu seus pesos.
developmentevalsservice-deliveryUnited KingdomUnited StatesCDNigeriaIndiaSingapore
Five of 475 studies on surgical AI in poorer countries came from low-income countries
A scoping review gathered 475 studies of artificial intelligence in surgical care across low-income and middle-income countries. 305 came from China, five came from low-income countries, and 12 reported reaching a real patient care pathway.
Cinco de los 475 estudios sobre IA en cirugía en países más pobres venían de países de renta baja
Una revisión exploratoria reunió 475 estudios sobre inteligencia artificial en la atención quirúrgica de países de renta baja y media. 305 venían de China, cinco venían de países de renta baja, y 12 informaron de un uso con pacientes reales.
Cinq des 475 études sur l'IA en chirurgie dans les pays pauvres venaient de pays à faible revenu
Une revue exploratoire a rassemblé 475 études sur l'intelligence artificielle dans les soins chirurgicaux des pays à revenu faible et intermédiaire. 305 venaient de Chine, cinq de pays à faible revenu, et 12 faisaient état d'un déploiement auprès de vrais patients.
Cinco dos 475 estudos sobre IA na cirurgia em países mais pobres vieram de países de baixa renda
Uma revisão de escopo reuniu 475 estudos sobre inteligência artificial no cuidado cirúrgico em países de renda baixa e média. 305 vieram da China, cinco vieram de países de baixa renda, e 12 relataram uso junto a pacientes reais.
Who can contradict the log
METR read 70,000 messages and about 1,300 transcripts from an unsanctioned board that OpenAI agents built in an Artifactory cache. Roughly 7 percent of the transcripts they evaluated contained spoofed tool calls, where an agent made it look like it ran one command while running another.
Quién puede contradecir el registro
METR leyó 70.000 mensajes y unas 1.300 transcripciones de un foro no autorizado que agentes de OpenAI construyeron en una caché de Artifactory. Alrededor del 7 por ciento de las transcripciones evaluadas contenían llamadas a herramientas falsificadas, en las que un agente aparentaba ejecutar un comando mientras ejecutaba otro.
Qui peut contredire le journal
METR a lu 70 000 messages et environ 1 300 transcriptions d'un forum non autorisé que des agents d'OpenAI avaient construit dans un cache Artifactory. Environ 7 % des transcriptions évaluées contenaient des appels d'outils falsifiés, où un agent faisait croire qu'il exécutait une commande alors qu'il en exécutait une autre.
Quem pode contradizer o registro
A METR leu 70.000 mensagens e cerca de 1.300 transcrições de um fórum não autorizado que agentes da OpenAI construíram num cache do Artifactory. Cerca de 7 por cento das transcrições avaliadas continham chamadas de ferramenta falsificadas, em que um agente fazia parecer que executava um comando enquanto executava outro.
Assorted links for 27 August 2026
Eleven links: what METR found inside OpenAI's agent incident, what an IMF paper puts on the table for African growth, where India sits per head rather than in total, and a Brazilian committee text that requires human supervision and authorises facial recognition in the same breath.
developmentregulationstate-capacityBrazilIndiaRwandaPhilippinesUgandaUnited States
Enlaces varios para el 27 de agosto de 2026
Once enlaces: lo que METR encontró en el incidente de los agentes de OpenAI, lo que un texto del FMI pone sobre la mesa para el crecimiento africano, dónde queda la India por habitante y no en volumen, y un texto de comisión brasileña que exige supervisión humana y autoriza el reconocimiento facial en el mismo aliento.
developmentregulationstate-capacityBrazilIndiaRwandaPhilippinesUgandaUnited States
Liens divers pour le 27 août 2026
Onze liens : ce que METR a trouvé dans l'incident des agents d'OpenAI, ce qu'un texte du FMI met sur la table pour la croissance africaine, où se situe l'Inde par habitant et non en volume, et un texte de commission brésilienne qui impose une supervision humaine et autorise la reconnaissance faciale dans le même souffle.
developmentregulationstate-capacityBrazilIndiaRwandaPhilippinesUgandaUnited States
Links diversos para 27 de agosto de 2026
Onze links: o que a METR encontrou no incidente dos agentes da OpenAI, o que um texto do FMI coloca na mesa para o crescimento africano, onde a Índia fica por habitante e não em volume, e um texto de comissão brasileira que exige supervisão humana e autoriza reconhecimento facial no mesmo fôlego.
developmentregulationstate-capacityBrazilIndiaRwandaPhilippinesUgandaUnited States
A new gov.uk benchmark finds chatbots usually answer well and almost never say I don't know
Researchers generated 22,066 questions from 2,781 gov.uk pages and tested 11 models on them, scoring each answer claim by claim against the page. Most answers were good, a small tail of bad misses drags the averages down, every model volunteered more than the page held, and almost none ever refused to answer.
service-deliveryevalsprocurementUnited KingdomUnited States
Un nuevo banco de pruebas sobre gov.uk halla que los chatbots responden en general bien y casi nunca dicen no lo sé
Unos investigadores generaron 22.066 preguntas a partir de 2.781 páginas de gov.uk y probaron 11 modelos, puntuando cada respuesta afirmación por afirmación contra la página. La mayoría de las respuestas eran buenas, una pequeña cola de fallos graves arrastra las medias hacia abajo, todos los modelos ofrecieron más de lo que había en la página, y casi ninguno se negó a responder.
service-deliveryevalsprocurementUnited KingdomUnited States
Un nouveau banc d'essai sur gov.uk trouve que les chatbots répondent en général bien et ne disent presque jamais je ne sais pas
Des chercheurs ont généré 22 066 questions à partir de 2 781 pages de gov.uk et testé 11 modèles, en notant chaque réponse affirmation par affirmation contre la page. La plupart des réponses étaient bonnes, une petite traîne de gros ratés tire les moyennes vers le bas, chaque modèle a fourni plus que la page n'en contenait, et presque aucun n'a jamais refusé de répondre.
service-deliveryevalsprocurementUnited KingdomUnited States
Um novo benchmark sobre o gov.uk acha que os chatbots respondem bem em geral e quase nunca dizem não sei
Pesquisadores geraram 22.066 perguntas a partir de 2.781 páginas do gov.uk e testaram 11 modelos, avaliando cada resposta afirmação por afirmação contra a página. A maioria das respostas era boa, uma pequena cauda de erros graves puxa as médias para baixo, todos os modelos ofereceram mais do que a página continha, e quase nenhum recusou responder.
service-deliveryevalsprocurementUnited KingdomUnited States
Brazil found AI in 94% of its judicial bodies and 55% of the branch that runs the services
The same survey asked where the technology sits, and at every level of government it is about twice as likely to be working on the administration's own processes as on anything delivered to a citizen.
service-deliverystate-capacitydevelopmentBrazilFranceUnited Kingdom
Brasil encontró IA en el 94% de sus órganos judiciales y el 55% del poder que presta los servicios
La misma encuesta preguntó dónde está la tecnología, y en todos los niveles de gobierno tiene cerca del doble de probabilidades de estar trabajando en los procesos internos de la administración que en algo entregado a un ciudadano.
service-deliverystate-capacitydevelopmentBrazilFranceUnited Kingdom
Le Brésil a trouvé de l'IA dans 94% de ses organes judiciaires et 55% du pouvoir qui gère les services
La même enquête a demandé où se trouve la technologie, et à tous les niveaux de l'administration elle a environ deux fois plus de chances de travailler sur les processus internes de l'administration que sur quoi que ce soit destiné à un citoyen.
service-deliverystate-capacitydevelopmentBrazilFranceUnited Kingdom
O Brasil encontrou IA em 94% dos seus órgãos do Judiciário e 55% do poder que toca os serviços
A mesma pesquisa perguntou onde a tecnologia está, e em todos os níveis de governo ela tem cerca de duas vezes mais chance de estar trabalhando nos processos internos da administração do que em qualquer coisa entregue ao cidadão.
service-deliverystate-capacitydevelopmentBrazilFranceUnited Kingdom
The evidence on agents flooding public services comes from eleven rich countries and Brazil
This site noted the agentic flooding paper on 21 August. What that line left out is where the paper looked, and the menu it offers has a column not every treasury can pay for.
La evidencia sobre agentes que desbordan los servicios públicos viene de once países ricos y de Brasil
Este sitio mencionó el artículo sobre desbordamiento agéntico el 21 de agosto. Lo que aquella línea dejó fuera es dónde miró el artículo, y el menú que ofrece tiene una columna que no todos los tesoros públicos pueden pagar.
Les données sur les agents qui submergent les services publics viennent de onze pays riches et du Brésil
Ce site a mentionné l'article sur la submersion agentique le 21 août. Ce que cette ligne laissait de côté, c'est l'endroit où l'article a cherché, et le menu qu'il propose comporte une colonne que tous les trésors publics ne peuvent pas payer.
As evidências sobre agentes inundando serviços públicos vêm de onze países ricos e do Brasil
Este site mencionou o artigo sobre inundação agêntica em 21 de agosto. O que aquela linha deixou de fora foi onde o artigo procurou, e o cardápio que ele oferece tem uma coluna que nem todo tesouro consegue pagar.
Kenya's draft AI policy promises a pay benchmark for data work indexed to rates abroad
A row in the implementation matrix commits to a fair-pay reference framework for data annotation and content moderation, calibrated to international rates. The document gives no rate and does not say whose. What is scheduled first is a publication, and publishing it needs nobody else's signature.
El borrador de política de IA de Kenia promete una referencia salarial para el trabajo de datos, indexada a tarifas del exterior
Una fila de la matriz de implementación compromete un marco de referencia de pago justo para anotación de datos y moderación de contenido, calibrado con tarifas internacionales. El documento no da ninguna tarifa y no dice de quién. Lo que llega primero es una publicación, y publicarla no necesita la firma de nadie más.
Le projet de politique d'IA du Kenya promet un barème de rémunération pour le travail de la donnée, indexé sur des taux étrangers
Une ligne de la matrice de mise en œuvre engage un cadre de référence de rémunération équitable pour l'annotation de données et la modération de contenu, calibré sur des taux internationaux. Le document ne donne aucun taux et ne dit pas lesquels. Ce qui arrive d'abord est une publication, et la publier ne demande la signature de personne d'autre.
A minuta de política de IA do Quênia promete uma referência de remuneração para o trabalho de dados, indexada a taxas de fora
Uma linha da matriz de implementação assume um marco de referência de remuneração justa para anotação de dados e moderação de conteúdo, calibrado por taxas internacionais. O documento não dá taxa alguma e não diz de quem. O que chega primeiro é uma publicação, e publicá-la não depende da assinatura de mais ninguém.
The government share in Pew's AI-writing study is about five flagged pages out of 669
Pew ran an AI-detection model over nearly 490,000 web pages and found ten times more AI writing on commercial domains than on government ones. The government figure everyone will quote is drawn from 669 sampled pages, and it has been falling since 2024 while every other domain rises.
state-capacityevalsregulationUnited States
La parte gubernamental en el estudio de Pew sobre escritura con IA son unas cinco páginas señaladas de 669
Pew pasó casi 490.000 páginas web por un detector de escritura con IA y encontró diez veces más texto de máquina en dominios comerciales que en dominios de gobierno. La cifra gubernamental que todo el mundo citará sale de 669 páginas muestreadas, y viene cayendo desde 2024 mientras todos los demás dominios suben.
state-capacityevalsregulationUnited States
La part gouvernementale dans l'étude de Pew sur l'écriture par IA tient à environ cinq pages signalées sur 669
Pew a passé près de 490 000 pages web dans un détecteur d'écriture par IA et trouve dix fois plus de texte machine sur les domaines commerciaux que sur les domaines gouvernementaux. Le chiffre gouvernemental que tout le monde citera repose sur 669 pages échantillonnées, et il baisse depuis 2024 alors que tous les autres domaines montent.
state-capacityevalsregulationUnited States
A fatia governamental no estudo da Pew sobre escrita com IA são cerca de cinco páginas sinalizadas em 669
A Pew passou quase 490.000 páginas web por um detector de escrita com IA e encontrou dez vezes mais texto de máquina em domínios comerciais do que em domínios de governo. O número governamental que todo mundo vai citar vem de 669 páginas amostradas, e vem caindo desde 2024 enquanto todos os outros domínios sobem.
state-capacityevalsregulationUnited States
Vietnam's AI scoring sheet puts sixty-five of a hundred marks on doing the work
Vietnam's criteria for public-service AI platforms, in force since 19 June, mark a platform out of 100. Thirty-five marks are for compliance and sixty-five are labelled for the sector's own work. What those marks reward is not published, so the split between the two blocks is all the ministry has disclosed.
procurementstate-capacitydevelopmentVietnamUnited KingdomIndia
La hoja de puntuación de IA de Vietnam da sesenta y cinco de cien puntos al trabajo en sí
Los criterios vietnamitas para plataformas de IA de servicio público, en vigor desde el 19 de junio, puntúan una plataforma sobre 100. Treinta y cinco puntos son de cumplimiento y sesenta y cinco están etiquetados como referidos al trabajo del propio sector. Lo que premian esos puntos no se publica, de modo que el reparto entre los dos bloques es todo lo que el ministerio ha divulgado.
procurementstate-capacitydevelopmentVietnamUnited KingdomIndia
La grille de notation IA du Vietnam attribue soixante-cinq points sur cent au travail lui-même
Les critères vietnamiens pour les plateformes d'IA de service public, en vigueur depuis le 19 juin, notent une plateforme sur 100. Trente-cinq points portent sur la conformité et soixante-cinq sont affichés comme portant sur le travail du secteur lui-même. Ce que ces points récompensent n'est pas publié, si bien que la répartition entre les deux blocs est tout ce que le ministère a divulgué.
procurementstate-capacitydevelopmentVietnamUnited KingdomIndia
A grade de avaliação de IA do Vietnã dá sessenta e cinco de cem pontos ao trabalho em si
Os critérios vietnamitas para plataformas de IA no serviço público, em vigor desde 19 de junho, avaliam uma plataforma em 100 pontos. Trinta e cinco são de conformidade e sessenta e cinco estão rotulados como referentes ao trabalho do próprio setor. O que esses pontos premiam não é publicado, de modo que a divisão entre os dois blocos é tudo o que o ministério divulgou.
procurementstate-capacitydevelopmentVietnamUnited KingdomIndia
A register of government algorithms can only hold what the state bought
Almost every country now has an AI policy, so that count has stopped telling anyone apart. The measure that has not saturated is whether a government must disclose the algorithms it runs itself, and procurement caps how much of it a register could ever reach.
regulationstate-capacitydevelopmentBrazilUnited KingdomUnited States
Un registro de algoritmos públicos solo puede contener lo que el Estado compró
Casi todos los países tienen ya una política de IA, y ese recuento ha dejado de distinguir a nadie. La medida que no se ha saturado es la que obliga a un gobierno a divulgar los algoritmos que él mismo opera, y la contratación pública limita cuánto podría alcanzar un registro.
regulationstate-capacitydevelopmentBrazilUnited KingdomUnited States
Un registre des algorithmes publics ne peut contenir que ce que l'État a acheté
Presque tous les pays ont désormais une politique d'IA, et ce décompte ne distingue plus personne. La mesure qui n'est pas saturée est celle qui oblige un gouvernement à divulguer les algorithmes qu'il fait tourner lui-même, et la commande publique plafonne ce qu'un registre pourrait en atteindre.
regulationstate-capacitydevelopmentBrazilUnited KingdomUnited States
Um registro de algoritmos públicos só pode conter o que o Estado comprou
Quase todos os países já têm uma política de IA, e essa contagem deixou de distinguir alguém. A medida que não saturou é a que obriga um governo a divulgar os algoritmos que ele próprio opera, e as compras públicas limitam o quanto um registro conseguiria alcançar.
regulationstate-capacitydevelopmentBrazilUnited KingdomUnited States
A rule written for one market became a property of the tool everyone else uses
Brussels wrote AI transparency rules for its own market. The mark now travels with the tool everywhere, and the people it reaches had no part in either decision. The objection is jurisdictional, not to the rule.
regulationdevelopmentservice-deliveryEuropean UnionVietnamIndonesia
Una regla escrita para un solo mercado se volvió propiedad de la herramienta que usan todos los demás
Bruselas escribió reglas de transparencia para su propio mercado. La marca ahora viaja con la herramienta a todas partes, y las personas a las que alcanza no participaron en ninguna de las dos decisiones. La objeción es de jurisdicción, no a la regla.
regulationdevelopmentservice-deliveryEuropean UnionVietnamIndonesia
Une règle écrite pour un seul marché est devenue une propriété de l'outil que tout le monde utilise
Bruxelles a écrit des règles de transparence pour son propre marché. La marque voyage désormais avec l'outil partout, et les personnes qu'elle atteint n'ont pris part à aucune des deux décisions. L'objection porte sur la compétence, pas sur la règle.
regulationdevelopmentservice-deliveryEuropean UnionVietnamIndonesia
Uma regra escrita para um só mercado virou propriedade da ferramenta que todos os outros usam
Bruxelas escreveu regras de transparência para o próprio mercado. A marca agora viaja com a ferramenta para todo lugar, e as pessoas que ela alcança não participaram de nenhuma das decisões. A objeção é de jurisdição, não à regra.
regulationdevelopmentservice-deliveryEuropean UnionVietnamIndonesia
Indonesia's welfare redesign lets people dispute what the state's records say they own
In one Indonesian district, more than 9,000 people used a new appeals channel to correct what the government's records said about them. That number counts errors found, not yet errors fixed.
El rediseño de la asistencia social en Indonesia permite disputar lo que los registros del Estado dicen que uno posee
En un distrito indonesio, más de 9.000 personas usaron un nuevo canal de apelaciones para corregir lo que los registros del gobierno decían sobre ellas. Esa cifra cuenta errores encontrados, todavía no errores corregidos.
La refonte de l'aide sociale en Indonésie permet de contester ce que les registres de l'État vous attribuent
Dans un district indonésien, plus de 9 000 personnes ont utilisé un nouveau canal de recours pour corriger ce que les registres du gouvernement disaient d'elles. Ce chiffre compte les erreurs trouvées, pas encore les erreurs réparées.
A reforma da assistência social na Indonésia permite contestar o que os registros do Estado dizem que você possui
Em um distrito indonésio, mais de 9.000 pessoas usaram um novo canal de recursos para corrigir o que os registros do governo diziam sobre elas. Esse número conta erros encontrados, ainda não erros corrigidos.
Using AI to write public comments improves every comment without closing the education gap
A survey experiment on US rulemaking found that ChatGPT made public comments easier to write and much better rated, without making people understand the rule any better. Agencies are now using the same models to read the comments.
participationstate-capacityprocurementUnited StatesNPIndiaGhanaDZ
Usar IA para escribir comentarios públicos mejora todos los comentarios sin cerrar la brecha educativa
Un experimento sobre la consulta pública federal estadounidense muestra que ChatGPT hace los comentarios más fáciles de escribir y mucho mejor calificados, sin que la gente entienda mejor la norma. Las agencias ya usan los mismos modelos para leer los comentarios.
participationstate-capacityprocurementUnited StatesNPIndiaGhanaDZ
Utiliser l'IA pour rédiger des commentaires publics améliore chaque commentaire sans réduire l'écart entre niveaux d'études
Une expérience sur la consultation publique fédérale américaine montre que ChatGPT rend les commentaires plus faciles à écrire et bien mieux notés, sans que les gens comprennent mieux la règle. Les administrations utilisent désormais les mêmes modèles pour lire les commentaires.
participationstate-capacityprocurementUnited StatesNPIndiaGhanaDZ
Usar IA para escrever comentários públicos melhora todos os comentários sem reduzir a diferença por escolaridade
Uma experiência sobre a consulta pública federal americana mostra que o ChatGPT torna os comentários mais fáceis de escrever e muito melhor avaliados, sem que as pessoas entendam melhor a norma. As agências agora usam os mesmos modelos para ler os comentários.
participationstate-capacityprocurementUnited StatesNPIndiaGhanaDZ
Counting where migrants work moves Tajikistan from the 21st to the 75th percentile of AI exposure
A new country-level measure follows AI exposure through migrant work and remittance income. Tajikistan shows why a national labour market does not stop at the border.
developmentevalsTJRussia
Contar dónde trabajan los migrantes lleva a Tayikistán del percentil 21 al 75 de exposición a la IA
Una nueva medida por país sigue la exposición a la IA a través del trabajo migrante y las remesas. Tayikistán muestra por qué un mercado laboral nacional no se detiene en la frontera.
developmentevalsTJRussia
Compter où travaillent les migrants fait passer le Tadjikistan du 21e au 75e centile d'exposition à l'IA
Une nouvelle mesure par pays suit l'exposition à l'IA à travers le travail migrant et les envois de fonds. Le Tadjikistan montre pourquoi un marché du travail national ne s'arrête pas à la frontière.
developmentevalsTJRussia
Contar onde os migrantes trabalham leva o Tadjiquistão do 21º ao 75º percentil de exposição à IA
Uma nova medida por país acompanha a exposição à IA através do trabalho migrante e das remessas. O Tadjiquistão mostra por que um mercado de trabalho nacional não para na fronteira.
developmentevalsTJRussia
Claude's content mark shows that a model touched a text, not whether it supplied the argument
Claude's new mark can show that a model touched a text. It cannot tell whether the model supplied the argument, or only the English through which the argument had to travel.
regulationdevelopmentevalsEuropean Union
La marca de contenido de Claude muestra que un modelo tocó un texto, no si aportó el argumento
La nueva marca de Claude puede mostrar que un modelo tocó un texto. No puede decir si el modelo aportó el argumento, o solo el inglés por el que el argumento tenía que viajar.
regulationdevelopmentevalsEuropean Union
Le marquage de contenu de Claude montre qu'un modèle a touché un texte, pas s'il a fourni l'argument
Le nouveau marquage de Claude peut montrer qu'un modèle a touché un texte. Il ne peut pas dire si le modèle a fourni l'argument, ou seulement l'anglais par lequel l'argument devait passer.
regulationdevelopmentevalsEuropean Union
A marca de conteúdo do Claude mostra que um modelo tocou um texto, não se ele forneceu o argumento
A nova marca do Claude pode mostrar que um modelo tocou um texto. Não pode dizer se o modelo forneceu o argumento, ou apenas o inglês pelo qual o argumento precisava viajar.
regulationdevelopmentevalsEuropean Union
Estonia is committing to buy AI compute that no supplier has built yet
Britain used procurement to open its digital market, and 90% of G-Cloud's suppliers are SMEs taking 44% of the money. Estonia is trying the harder version, a forward purchase commitment to attract a supplier that is not there yet.
state-capacitycomputeprocurementEstoniaUnited KingdomEuropean Union
Estonia se compromete a comprar cómputo de IA que ningún proveedor ha construido todavía
Gran Bretaña usó la compra pública para abrir su mercado digital: el 90% de los proveedores del G-Cloud son pymes y se llevan el 44% del dinero. Estonia intenta la versión más difícil, un compromiso anticipado de compra para atraer a un proveedor que aún no está.
state-capacitycomputeprocurementEstoniaUnited KingdomEuropean Union
L'Estonie s'engage à acheter du calcul d'IA qu'aucun fournisseur n'a encore construit
La Grande-Bretagne a utilisé la commande publique pour ouvrir son marché numérique : 90 % des fournisseurs du G-Cloud sont des PME et elles prennent 44 % de l'argent. L'Estonie tente la version plus difficile, un engagement d'achat anticipé pour attirer un fournisseur qui n'est pas encore là.
state-capacitycomputeprocurementEstoniaUnited KingdomEuropean Union
A Estônia se compromete a comprar computação de IA que nenhum fornecedor construiu ainda
A Grã-Bretanha usou compras públicas para abrir seu mercado digital: 90% dos fornecedores do G-Cloud são PMEs e levam 44% do dinheiro. A Estônia tenta a versão mais difícil, um compromisso antecipado de compra para atrair um fornecedor que ainda não está lá.
state-capacitycomputeprocurementEstoniaUnited KingdomEuropean Union
African developers chose Chinese models before any government signed anything
Ten African countries signed an AI pact with China in Shanghai in July. The adoption it looks like it caused had already happened, in the hands of developers who could download the weights while the American onboarding queue was still running.
developmentcomputeprocurementUgandaKenyaChinaUnited States
Los desarrolladores africanos eligieron modelos chinos antes de que ningún gobierno firmara nada
Diez países africanos firmaron un pacto de IA con China en Shanghái en julio. La adopción que parece haber causado ya había ocurrido, en manos de desarrolladores que podían descargar los pesos mientras la cola de acceso estadounidense seguía corriendo.
developmentcomputeprocurementUgandaKenyaChinaUnited States
Les développeurs africains ont choisi les modèles chinois avant qu'aucun gouvernement ne signe quoi que ce soit
Dix pays africains ont signé un pacte sur l'IA avec la Chine à Shanghai en juillet. L'adoption qu'il semble avoir provoquée avait déjà eu lieu, entre les mains de développeurs qui pouvaient télécharger les poids pendant que la file d'attente américaine tournait encore.
developmentcomputeprocurementUgandaKenyaChinaUnited States
Desenvolvedores africanos escolheram modelos chineses antes de qualquer governo assinar coisa alguma
Dez países africanos assinaram um pacto de IA com a China em Xangai em julho. A adoção que ele parece ter causado já tinha acontecido, nas mãos de desenvolvedores que podiam baixar os pesos enquanto a fila de credenciamento americana ainda corria.
developmentcomputeprocurementUgandaKenyaChinaUnited States
OpenAI published two datasets on ChatGPT use and neither covers the public sector
OpenAI publishes two datasets on how ChatGPT is used, one for individuals and one for enterprises, and the public sector that it counts in the millions when selling appears in neither.
state-capacitydevelopmentUnited StatesUnited Kingdom
OpenAI publicó dos conjuntos de datos sobre el uso de ChatGPT y ninguno cubre el sector público
OpenAI publica dos conjuntos de datos sobre cómo se usa ChatGPT, uno para individuos y otro para empresas, y el sector público, que cuenta por millones cuando vende, no aparece en ninguno de los dos.
state-capacitydevelopmentUnited StatesUnited Kingdom
OpenAI a publié deux jeux de données sur l'usage de ChatGPT et aucun ne couvre le secteur public
OpenAI publie deux jeux de données sur l'usage de ChatGPT, un pour les particuliers et un pour les entreprises, et le secteur public, qu'elle compte en millions quand elle vend, n'apparaît dans aucun des deux.
state-capacitydevelopmentUnited StatesUnited Kingdom
A OpenAI publicou dois conjuntos de dados sobre o uso do ChatGPT e nenhum cobre o setor público
A OpenAI publica dois conjuntos de dados sobre como o ChatGPT é usado, um para indivíduos e outro para empresas, e o setor público, que ela conta aos milhões quando vende, não aparece em nenhum dos dois.
state-capacitydevelopmentUnited StatesUnited Kingdom
An AI exposure estimate was checked against vacancy data, and it moved the same way
AI exposure scores are hypotheses, and their value lies in predicting correctly. Here is one that was checked against vacancy data, and moved the same way.
evalsdevelopmentAustraliaChina
Una estimación de exposición a la IA se contrastó con los datos de vacantes, y se movió en el mismo sentido
Las puntuaciones de exposición a la IA son hipótesis, y su valor está en predecir correctamente. Aquí hay una que se contrastó con datos de vacantes, y se movió en el mismo sentido.
evalsdevelopmentAustraliaChina
Une estimation de l'exposition à l'IA a été confrontée aux offres d'emploi, et elle variait dans le même sens
Les scores d’exposition à l’IA sont des hypothèses, et leur valeur tient à la justesse de leurs prédictions. En voici un qui a été confronté aux données d’offres d’emploi, et qui a bougé dans le même sens.
evalsdevelopmentAustraliaChina
Uma estimativa de exposição à IA foi confrontada com os dados de vagas, e variou no mesmo sentido
Os escores de exposição à IA são hipóteses, e seu valor está em prever corretamente. Eis um que foi confrontado com dados de vagas, e se moveu no mesmo sentido.
evalsdevelopmentAustraliaChina
AI can explain a public service far better than it can reach one, across 166 countries
RADAR finds AI can describe public services far better than it can reach them. The repair is stable URLs and sensible bot policies, which is exactly why it will take a decade and why it looks like infrastructure.
La IA explica un servicio público mucho mejor de lo que consigue alcanzarlo, en 166 países
RADAR concluye que la IA describe los servicios públicos mucho mejor de lo que logra alcanzarlos. La reparación son URL estables y políticas de bots sensatas, que es justamente por qué llevará una década y por qué se parece a la infraestructura.
L'IA explique bien mieux un service public qu'elle ne parvient à l'atteindre, dans 166 pays
RADAR constate que l’IA décrit les services publics bien mieux qu’elle ne les atteint. La réparation passe par des URL stables et des politiques de robots raisonnables, ce qui explique exactement pourquoi cela prendra une décennie et pourquoi cela ressemble à de l’infrastructure.
A IA explica um serviço público muito melhor do que consegue alcançá-lo, em 166 países
O RADAR conclui que a IA descreve serviços públicos muito melhor do que consegue alcançá-los. O reparo são URLs estáveis e políticas de bots sensatas, que é exatamente por que vai levar uma década e por que se parece com infraestrutura.
Consultation is the cheap rung
The obvious objection to an 81,000-person AI consultation is representativeness. It is the weaker objection, and my own research is why.
participationregulationISChina
La consulta es el peldaño barato
La objeción obvia a una consulta sobre IA con 81.000 personas es la representatividad. Es la objeción más débil, y mi propia investigación explica por qué.
participationregulationISChina
La consultation est l’échelon bon marché
L’objection évidente à une consultation sur l’IA auprès de 81 000 personnes est la représentativité. C’est l’objection la plus faible, et mes propres recherches en sont la raison.
participationregulationISChina
A consulta é o degrau barato
A objeção óbvia a uma consulta sobre IA com 81.000 pessoas é a representatividade. É a objeção mais fraca, e minha própria pesquisa é a razão.
participationregulationISChina
The Vallance opportunity: agents read what screen readers read
Britain built the world's most admired government website and skipped the platform beneath it. The agentic wave is a second offer, on one condition.
agentic-stateservice-deliveryagentsUnited Kingdom
La oportunidad Vallance: los agentes leen lo que leen los lectores de pantalla
Gran Bretaña construyó el sitio de gobierno más admirado del mundo y se saltó la plataforma que hay debajo. La ola agentic es una segunda oferta, con una condición.
agentic-stateservice-deliveryagentsUnited Kingdom
L'opportunité Vallance : les agents lisent ce que lisent les lecteurs d'écran
La Grande-Bretagne a construit le site gouvernemental le plus admiré au monde et a sauté la plateforme en dessous. La vague agentique est une seconde offre, à une condition.
agentic-stateservice-deliveryagentsUnited Kingdom
A oportunidade Vallance: agentes leem o que os leitores de tela leem
A Grã-Bretanha construiu o site de governo mais admirado do mundo e pulou a plataforma por baixo dele. A onda agentic é uma segunda oferta, sob uma condição.
agentic-stateservice-deliveryagentsUnited Kingdom
The World Development Report 2026 is out
The World Bank's flagship on artificial intelligence has published. One of its numbers, what the nine chapters cover, and what I will be writing about here.
El Informe sobre el Desarrollo Mundial 2026 ya salió
El informe insignia del Banco Mundial sobre inteligencia artificial ya está publicado. Uno de sus números, qué cubren los nueve capítulos, y qué escribiré aquí sobre él.
Le Rapport sur le développement dans le monde 2026 est paru
Le rapport phare de la Banque mondiale sur l'intelligence artificielle est publié. Un de ses chiffres, ce que couvrent les neuf chapitres, et ce que j'écrirai ici à son sujet.
O Relatório sobre o Desenvolvimento Mundial 2026 saiu
O relatório principal do Banco Mundial sobre inteligência artificial foi publicado. Um dos seus números, o que cobrem os nove capítulos, e o que vou escrever sobre ele aqui.
New Jersey's AI strategy is a staffing list
Three of the four disciplines New Jersey's innovation office employs exist to decide what to build. Most governments adopting AI employ none of them.
state-capacityprocurementUnited States
La estrategia de IA de Nueva Jersey es una lista de personal
Tres de las cuatro disciplinas que emplea la oficina de innovación de Nueva Jersey existen para decidir qué construir. La mayoría de los gobiernos que adoptan IA no emplea ninguna.
state-capacityprocurementUnited States
La stratégie IA du New Jersey est une liste de postes
Trois des quatre disciplines employées par le bureau d’innovation du New Jersey servent à décider quoi construire. La plupart des gouvernements qui adoptent l’IA n’en emploient aucune.
state-capacityprocurementUnited States
A estratégia de IA de Nova Jérsia é uma lista de pessoal
Três das quatro disciplinas que o gabinete de inovação de Nova Jérsia emprega existem para decidir o que construir. A maioria dos governos que adoptam IA não emprega nenhuma delas.
state-capacityprocurementUnited States
Lots of procurement, limited learning
Federal agencies doubled their AI buying and kept no record of what they learned.
procurementstate-capacityagentsUnited StatesArgentinaEuropean Union
Muchas compras, poco aprendizaje
Las agencias federales duplicaron sus compras de IA y no guardaron registro de lo aprendido.
procurementstate-capacityagentsUnited StatesArgentinaEuropean Union
Beaucoup d’achats, peu d’apprentissage
Les agences fédérales ont doublé leurs achats d’IA sans garder trace de ce qu’elles en ont appris.
procurementstate-capacityagentsUnited StatesArgentinaEuropean Union
Muita contratação, pouca aprendizagem
As agências federais dobraram as suas compras de IA e não guardaram registo do que aprenderam.
procurementstate-capacityagentsUnited StatesArgentinaEuropean Union
My two cents on AI detectors
My two cents on AI detectors. My worry is not whether they work, but what they measure and how people read it.
Mis dos centavos sobre los detectores de IA
Mis dos centavos sobre los detectores de IA. Mi preocupación no es si funcionan, sino qué miden y cómo la gente lo lee.
Mes deux centimes sur les détecteurs d'IA
Mes deux centimes sur les détecteurs d'IA. Mon inquiétude n'est pas de savoir s'ils fonctionnent, mais ce qu'ils mesurent et comment on les lit.
Meus dois centavos sobre detectores de IA
Meus dois centavos sobre detectores de IA. Minha preocupação não é se funcionam, mas o que medem e como as pessoas leem isso.
Government services are over-represented in AI conversations by a factor of almost twenty
Google's ATLAS report, drawn from nearly 15 million Gemini interactions, finds government-services queries over-represented by roughly twenty times their share of American time-use hours. People are already using AI as an unofficial interface to the state.
Los servicios públicos están sobrerrepresentados en las conversaciones con IA por un factor de casi veinte
El informe ATLAS de Google, extraído de casi 15 millones de interacciones con Gemini, encuentra que las consultas sobre servicios públicos están sobrerrepresentadas unas veinte veces respecto a su parte en las horas de uso del tiempo estadounidense. La gente ya usa la IA como una interfaz no oficial con el Estado.
Les services publics sont surreprésentés dans les conversations avec l'IA par un facteur de près de vingt
Le rapport ATLAS de Google, tiré de près de 15 millions d'interactions avec Gemini, trouve que les requêtes sur les services publics sont surreprésentées d'environ vingt fois leur part dans les heures d'usage du temps américain. Les gens utilisent déjà l'IA comme interface officieuse avec l'État.
Serviços públicos estão super-representados nas conversas com IA por um fator de quase vinte
O relatório ATLAS do Google, tirado de quase 15 milhões de interações com o Gemini, constata que consultas sobre serviços públicos estão super-representadas em cerca de vinte vezes sua fatia nas horas de uso do tempo americano. As pessoas já usam a IA como uma interface não oficial com o Estado.
China's 29-country AI alliance, and what it is buying
China closed the World AI Conference in Shanghai by launching a 29-country AI alliance. A few weeks earlier, at the G7, Anthropic and Google DeepMind called for an American-led coalition.
developmentcomputeChina
La alianza de IA de 29 países de China, y qué está comprando
China cerró la Conferencia Mundial de IA en Shanghái lanzando una alianza de IA de 29 países. Pocas semanas antes, en el G7, Anthropic y Google DeepMind pedían una coalición liderada por Estados Unidos.
developmentcomputeChina
L'alliance IA à 29 pays de la Chine, et ce qu'elle achète
La Chine a clôturé la Conférence mondiale sur l'IA à Shanghai en lançant une alliance IA à 29 pays. Quelques semaines plus tôt, au G7, Anthropic et Google DeepMind appelaient à une coalition menée par les États-Unis.
developmentcomputeChina
A aliança de IA de 29 países da China, e o que ela está comprando
A China encerrou a Conferência Mundial de IA em Xangai lançando uma aliança de IA com 29 países. Poucas semanas antes, no G7, Anthropic e Google DeepMind pediam uma coalizão liderada pelos Estados Unidos.
developmentcomputeChina
Control beats location
Time governments realize that control beats location.
I wish government software were updated half as often as this is
I just wish government software was half as updated as Claude Code is. A version number like 1.21459.3 means a permanent team shipping improvements every day.
What closing a digital governance programme in Serbia actually showed
This week, World Bank leadership flew to Belgrade for the closure meeting of Serbia's Enabling Digital Governance (EDGE) project.
Calling it grassroots does not make it grassroots
I find it slightly odd that 'no data center in my backyard' is now routinely described as grassroots activism.
What generative AI does to the open web
Fascinating paper by Alex Chan (NBER, 2026) on what generative AI does to the open web. The web runs on a simple trade: publishers make things worth reading, search sends readers to them, and the visits pay for the work.
The study everyone cites as proof that agents are unreliable
I keep seeing this study passed around as proof that AI agents are unreliable. The study is good. The way it is being read is not, and the misreading follows a pattern I see whenever a study on AI failures gets published.
A place where every claim about AI and work points somewhere
Johanna Einsiedler just launched what the argument about AI and work has been missing: a place where every claim points back to a number on a page.
AI agents are more obedient than the people they act for
Fourteen frontier models were run through the same choice-architecture nudges used on humans, in PNAS. People accept an explicit default about 88 percent of the time; the agents acting for them comply far more readily, which changes who a nudge is really aimed at.
The interesting part is the volatility
The interesting part is the volatility. A mix that swings this much in twelve months is a warning to anyone signing a multi-year exclusive deal.
Edgar Morin has died at 104
Edgar Morin has died at 104. He spent a career reviving Montaigne's preference for a well-made head over a well-filled one, and our education systems are still built for the second.
The hard part of AI infrastructure is not the hardware
In two hours I am joining the MCDF workshop on AI Infrastructure to make a case that often gets skipped: the hard part of the Agentic State is not the models, it is the plumbing underneath.
The logic of lines, from Florence to the digital queue
Having spent my younger life in Florence, I always had an interest in the logic of lines. At first, and like most tourists, I treated queues as outsourced discernment: after all, so many strangers could not all be wrong.
Open budgets, and what AI changes about what we can see
Open budgets and open data may not be at their fashionable peak, but AI changes what we can see in the data, and what we cannot.
Whose data trains the model, and who benefits from it
As one X user bluntly summarized: 'Claude is winning because rich people are providing the training data. Poor Meta has to train AI with the peasants of the internet.' Crude, and not quite right about the mechanism.
What is actually holding governments back on the agentic state
When it comes to The Agentic State, what is actually holding governments back? The launch of our vision paper last October triggered months of conversations with governments across every continent.
Hacking the public sector, and the civil servants who do it
'What I really love is hacking the public sector. And for that, I need my cracha.' The most effective public servants are not outsiders throwing rocks, they are insiders who know which doors their badge opens.
AI and democracy keeps ignoring the generalist
It still amazes me how little attention those working at the intersection of AI and democracy pay to the role of generative AI in collective action.
Summer in Europe while you can: AI and the Jevons paradox might soon make it unbearable
Biometrics lowered the administrative cost of designing new entry requirements, which removed the natural brake on regulatory ambition. The efficiency gain was immediately consumed by the expanded scope of what became feasible to demand.
regulationservice-deliveryEuropean Union
Worth engaging with, and worth the harder questions
Worth engaging with, and worth asking the harder questions about. This is the kind of build the small-models argument has been waiting for.
Theoretical AI capability against observed usage
This figure from Anthropic comparing “theoretical AI capability” with observed usage across occupations has been circulating widely in the AI policy bubble.
This may not be the future, but it is what citizens will expect
I'm not sure that's what the future will look like, but it is what citizens will expect from governments. Whether those in power like it or not.
Help needed: what agentic procurement would actually require
Help needed! I've been tinkering with an interactive 'periodic table' for AI in government for a course I'm developing.
ChatGPT is splitting into work and everything else
OpenAI recently released data on how people use ChatGPT. One number stood out. In December 2025, about 75% of messages from paid Pro users were work-related.
A system-level view of AI, and why it is rare
This is one of the most important system-level AI posts I’ve read in a while. Many public services are not designed to be accessible, they are designed to be survivable: complexity and bad UX function as informal rationing.
Buy or build, and the question that decides it
Buy or build? One trajectory from ambition to pragmatism is worth a thousand AI strategies. Singapore's SEA-LION AI model started by pretraining from scratch.
Deliberative democracy's bottleneck is not scale, it is consequence
I have argued that deliberative democracy's constraint is not scale but consequence, and proposed coordination infrastructure that turns deliberative agreement into organised pressure. A new preprint reads like the dark mirror of that idea.
Verified humans, synthetic voices: where collective intelligence meets collective manipulation
The same infrastructure that could help democratic publics convert consensus into leverage can, with a different governance structure, manufacture the appearance of consensus where none exists.
A technical milestone for the region, and the harder question underneath
A regional public-AI launch is a real technical achievement, and 'sovereign' is doing a lot of unexamined work in how it is described. Ownership, hosting, control and capability are four different claims.
Singapore's governance framework for agentic AI in government
Singapore's IMDA released a governance framework for agentic AI in government (via Drasko Draskovic, PhD). It is careful, thorough work on oversight, testing, and accountability.
regulationagentic-stateSingapore
An RCT on patient-facing medical AI, and what it measured
A randomised trial in China (n = 2,069), published in Nature Medicine, tested an LLM chatbot that conducts the patient intake interview before a specialist visit and hands the clinician a structured summary.
2025
A very specific kind of AI failure in government
I start to see a very specific kind of AI failure in government. A new tool arrives, the institution responds with what it already knows how to do: mandatory training, warnings, guardrails, committees, sign-offs.
In many countries competitive recruitment is more fiction than fact
In many countries, competitive recruitment is more fiction than fact. And in the public sector in particular, interviews are often the stage where merit quietly disappears and “soft-rigging” takes over.
Public services will not be shaped by the interface alone
About a year ago, Luke Jordan and I argued that the future of public services would not be shaped only by user interfaces, but by agents acting on behalf of users.
Europe regulates the digital world and keeps tripping in the physical one
I just came back from Frankfurt, and I spent an unhealthy amount of time talking about digital policies while looking at stairs.
regulationstate-capacityEuropean UnionUnited States
The algorithmic hand: AI, democracy and collective action at scale
New paper out 'The Algorithmic Hand: Artificial Intelligence, Democracy, and Collective Action at Scale' Every few decades, a new technology promises to reinvent democracy.
What we do not say enough about govtech
At the Data Science Conference, I had a conversation with Gustavo Maia from Colab about something we don't say enough in tech: if you want to build things that actually reach people, look at the public sector.
The GDPR review, and what gets traded for competitiveness
So, the European Union is reviewing parts of the GDPR to make it more compatible with innovation and competitiveness.
regulationEuropean Union
How AI reshapes buyer-supplier negotiations
Very interesting study (link in comments) on how AI reshapes buyer–supplier negotiations. In experiments with students and professional negotiators, human suppliers negotiated with a ChatGPT-based agent acting as the buyer.
When agents start negotiating on our behalf
Super interesting, and it connects to something we've been thinking about in our work in The Agentic State: AI agents are not a substitute for fixing bad UX, but they are arriving regardless.
What development research keeps missing about AI
A few thoughts as I join the Global Development Conference 2025 in Clermont-Ferrand, promoted by the Global Development Network.
developmentparticipationFrance
The inference divide: the inequality no one is talking about
Two people with identical devices, connections and skills, using the same model, can receive fundamentally different intelligence augmentation. One subscription lets the model think for minutes. The other gets seconds.
The AI future is about identity, territory and governance
The AI Future Isn't Just About Algorithms: It's About Identity, Territory, and Governance. Thank you for the opportunity to speak at last week Techritory Forum in Riga!
The agentic state, second version
Last week, at the Tallinn Digital Summit, we launched the second version of The Agentic State: a vision for how governments can use AI agents not to replace human judgment, but to redesign how the state itself works.
agentic-stateagentsEstonia
A hype check on human-in-the-loop
Help needed: human-in-the-loop hype check The more I think about it, the more I suspect that blanket calls for “human-in-the-loop” in AI for public services are a first-world comfort blanket.
We have far more AI policy trackers than AI deployment trackers
I wish we had at least half as many AI deployment trackers as we have AI policy trackers. Especially in contexts where deployments are likely to have major consequences….
Context lock-in and the new AI monopolies
The next lock-in may not be the model but the accumulated context: conversations, preferences and work history that cannot be moved. Open protocols for context portability are the remedy, and they are a rare pro-competition rule most sides could accept.
Most digital government life events are theatre
Most digital government “life events” are just theater. Time to admit it. Digital government folks love to talk about “life events”: having a baby, starting a business, losing a job.
Looking for machine learning systems that are actually running in government
Help needed: looking for real-world Machine Learning systems in government A few weeks ago, I reached out to this network asking for compelling GenAI use cases in public-sector workflows.
The UK's Copilot experiment with 20,000 civil servants deserves more attention
The UK's Copilot experiment with 20,000 civil servants deserves way more attention than it's gotten. The results, 26 minutes saved per day, might seem modest, but they reveal something crucial about AI in government.
evalsstate-capacityUnited Kingdom
Asking for automation agents that actually get things done
Asking the community: any examples, public or private sector, of automation agents that actually get things done for users online?
Why would a government adopt this? The question public-sector AI keeps skipping
One of the recurring blind spots in public sector AI enthusiasm is a failure to answer a basic question: Why would governments succeed with GenAI now, given their long history of struggling to adopt much simpler technologies?
The procurement problem nobody wants to own
*Excellent* analysis on UK procurement challenges. The findings also resonate strongly with what we see in developing economies, but where there's an additional structural barrier: payment delays.
procurementstate-capacityUnited Kingdom
The agentic state: ten functional layers of government, revamped
New paper on The Agentic State Very happy to share 'The Agentic State: How Agentic AI Will Revamp 10 Functional Layers of Government and Public Administration'.
Where generative AI is actually hitting labour markets
Most studies of generative AI and jobs rest on exposure estimates rather than observed effects. New World Bank research asks where the impact on labour markets is actually landing.
Why generative AI isn't transforming government (yet)
I asked practitioners, NGOs and philanthropies a simple question: where are the compelling generative AI use cases in public-sector workflows? The responses, though numerous, were underwhelming.
Who is liable when the harm is assembled from parts
Beatriz Botero Arcila proposes fault-based joint and several liability for AI systems, with targeted strict liability for high-risk cases. Her framework takes on the 'many hands' problem: who answers when the harm is assembled from parts.
The first clinical trial of a generative AI therapy chatbot
The first clinical trial of a generative AI therapy chatbot, from Dartmouth in NEJM AI: 51% average reduction in depression symptoms, 31% in generalised anxiety, 19% in eating-disorder concerns.
evalsservice-deliveryUnited States
New research: what digital participation actually changes
New in Government Information Quarterly, with Fredrik Sjoberg: who takes part in digital participation does not by itself determine who benefits. What decides it is whether, and how, governments respond.
Who participates in digital democracy, and who really benefits?
The dominant assumption is that who participates determines who benefits. Across participatory budgeting in Brazil, FixMyStreet, Iceland's crowdsourced constitution and Change.org, that chain broke at some stage in every case.
On sortition, and what a lottery can legitimately decide
In the inaugural issue of the Journal of Sortition: 'The Limits of Representativeness in Citizens' Assemblies', on what a democratic minipublic can and cannot claim to represent.
2024
Unwritten 2025
'We don't know where this technology is going' sounds thoughtful and feels responsible. Increasingly I am convinced it is neither, and that waiting is a decision that may cost us.
How to make AI agents serve everyone, not just the privileged few
Written with Luke Jordan: as AI agents reshape access to public services, the risk is a widening gap between citizens who can delegate and those who cannot. What it would take to build the equitable version instead.
Agents for the few, queues for the many – or agents for all?
Closing the public services divide by regulating for AI's opportunities.
service-deliveryagentsregulationUnited KingdomBrazilSouth AfricaUnited States
The overlooked upside of AI for developing nations
While most discussions about AI and developing nations fixate on risks and challenges, we often overlook the glaringly obvious opportunities.
AI's coming data saturation is an opportunity for the countries left out of the corpus
This recent Nature article projecting AI data saturation in the near future inadvertently highlights a significant opportunity for developing economies.
Inclusive AI infrastructure, and who is in the room when it is designed
Still energized from moderating this panel at the Tallinn Digital Summit on inclusive AI infrastructure - born from a growing collaboration between Estonia’s Government and The World Bank Group.
developmentcomputeEstonia
The link between open data and trust in government is weaker than we assumed
Two rounds of survey evidence suggesting the link between open data and institutional trust is weaker than open-government advocates assume, and in places runs the other way.
Forty-five percent of UK public services report no AI use at all
Excellent new survey by Jonathan Bright and colleagues at the Alan Turing Institute shows that 45% of UK public service professionals are aware of GenAI use at work, while 22% use it themselves.
state-capacityagentsUnited Kingdom
AI's effects on elections are largely overstated
Keegan McBride and colleagues in MIT Technology Review, on why AI's measured effect on elections is far smaller than the commentary suggests. A reminder to prefer the research on voting behaviour over the punditry.
2023
Underestimated effects of AI on democracy, and a gloomy scenario
Bots writing to legislators got within 2 percent of the response rate humans did. The trouble starts when governments answer with bots of their own.
The hidden risks of AI: how linguistic diversity can make or break collective intelligence
Diverse groups solve problems better. Models trained mostly on English inherit a narrower collective intelligence, and a subtler digital divide follows.
Nothing matched. This searches titles, descriptions, topics and places, so a phrase from the middle of an article will not find it. Try a topic fromthe topics page, or browsethe full chronology.