Weekly AI Watch

Every week, what changed and what it means for you. No hurry, no alarm.

Week of 20 November 2026 | Past editions15,959 papers found · 15,002 read · 5,264 belonged · 42 stories passed every check

Observatorio de la
Inteligencia Artificial

Cada semana, qué cambió y qué significa para usted. Sin prisa y sin alarma.

Semana del 20 de noviembre de 2026 | Ediciones anteriores15,959 estudios encontrados · 15,002 leídos · 5,264 correspondían · 42 notas pasaron todas las comprobaciones

weeklyAI · Week of 20 November 2026weeklyAI · Semana del 20 de noviembre de 2026

← Your vote← Su voto

other · Frontiers in Political Science · la publicación, 5 nov 2026 · gratis

Un protocolo para revisar si dos asistentes de inteligencia artificial del gobierno de Kazajistán responden lo mismo en kazajo y en ruso; todavía no hay resultados

Versión breve · la versión detallada sigue, unos 5 min

Pregunte al chat de weeklyAI

Pregúnteme sobre la IA en su vida: le señalo las notas que weeklyAI ha publicado sobre el tema, esta semana y antes.

Las conversaciones se guardan mientras exista weeklyAI, para mejorar la publicación. Se responde en el idioma en que usted escribe.

El estudio, de un vistazo
Quiénes
Dos asistentes de inteligencia artificial del gobierno de Kazajistán: uno de trámites y Sailau AI, el chatbot electoral
Cuántos
60 escenarios, 30 por cada interfaz
Dónde
Kazajistán
Cuándo
Protocolo publicado el 5 de noviembre de 2026
Tipo de estudio
analysis of what people did
Quién lo hizo
Universidad Nacional Pedagógica de Abai y Universidad Médica Nacional de Asfendiyarov, Kazajistán
El límite que importa
Es solo un plan: no se ha hecho ninguna pregunta ni se ha visto ninguna respuesta, así que no hay datos

Lo que el plan revisa en cada asistente

Respuesta de primer intento y sus enlacesfrente aPantalla de entrada y caminos fijos de ayuda, contacto, corrección, queja y privacidad

Se revisan por separado, en kazajo y en ruso, y en cada visita

Kazajofrente aRuso

Se compara si ambos idiomas conservan la misma guía decisiva frente a una referencia legal bloqueada

No se ha hecho ninguna pregunta ni se ha visto ninguna respuesta, así que no hay ningún dato sobre su exactitud.
Lectura de weeklyAI

1. El 5 de noviembre de 2026 se publicó un plan de revisión, firmado por investigadores de dos universidades de Kazajistán, para dos asistentes de inteligencia artificial que el gobierno de ese país enlaza desde sus páginas oficiales: uno de trámites y otro de información electoral, este último en Telegram desde el sitio de la comisión electoral. El plan describe cómo se les harían preguntas iguales en kazajo y en ruso y se compararían sus respuestas con la ley vigente. No se ha hecho ninguna pregunta ni se ha visto ninguna respuesta, así que no hay ningún dato sobre su exactitud.

2. El plan revisa dos cosas por separado. Primero mira la primera respuesta que da cada asistente y los enlaces que trae. Después mira la pantalla de entrada y los caminos fijos de ayuda, contacto, corrección, queja y privacidad que se alcanzan desde ahí. Anota todo esto en kazajo y en ruso, y en cada visita.

3. La comisión electoral informa que su asistente funciona en kazajo y en ruso, que usa fuentes de la ley electoral y que se basa en un modelo llamado AlemLLM. El texto solo cita la ley de Kazajistán: la Constitución, vigente desde el 1 de julio de 2026, establece el kazajo como idioma estatal y permite el uso oficial del ruso. El plan advierte que una respuesta de estos asistentes no tiene por sí misma valor legal.

Qué significa para usted

Para usted, que va a votar, esto todavía no dice si esos asistentes contestan bien o mal: el plan solo describe cómo se revisarían, y ninguna pregunta se ha hecho. Cuando el estudio se publique, conviene mirar si comparó las respuestas con la ley vigente en cada idioma y si encontró diferencias entre el kazajo y el ruso. Por ahora, si un asistente oficial le responde, tenga presente que esa respuesta no tiene valor legal por sí misma.

Tilep, B., Sultan, Y., Akaidar, K. et al. (2026). Auditing multilingual public information through officially linked AI interfaces in Kazakhstan. Frontiers in Political Science. https://doi.org/10.3389/fpos.2026.1984220

Quién pagó: Los autores declararon que no se recibió apoyo financiero para este trabajo ni para su publicación, y declararon no tener relaciones comerciales o financieras que pudieran constituir un conflicto de interés.

Los hallazgos de otros estudios que aquí se mencionan provienen de este documento; esos estudios no se consultaron uno por uno.

Versión detalladaLos pasajes copiados del artículo, las ilustraciones y cada fuente con cuánto leímos de ella · 1051 palabras · unos 5 minLeerla →Cerrar

La regla y la evidencia: un protocolo para revisar si dos asistentes de inteligencia artificial del gobierno de Kazajistán responden lo mismo en kazajo y en ruso; todavía no hay resultados

### EL ACTO

El acto, en este caso, no es una resolución ni una sentencia: es un protocolo de estudio, publicado el 5 de noviembre de 2026, que describe cómo se auditará una interfaz de inteligencia artificial del gobierno de Kazajistán y un chatbot electoral llamado Sailau AI. El protocolo pregunta si las respuestas de primer intento coinciden con una referencia legal bloqueada, si dan una guía materialmente equivalente en kazajo y en ruso, y si muestran rutas de verificación, corrección, revisión humana, queja y reparación1.

El texto no impone nada a nadie ni entra en vigor para ningún votante. Lo que establece es un método: un conjunto de 60 escenarios, 30 por cada interfaz, con dos formulaciones en cada idioma obligatorio, lo que da 240 administraciones de primer intento en kazajo y ruso, y 336 observaciones generales, con un máximo de 420 si se activa el inglés. Los escenarios se repiten a las 24 horas y a los 14 días.

Quién lo diseñó: el equipo de los autores, en la Universidad Nacional Pedagógica de Abai y la Universidad Médica Nacional de Asfendiyarov, ambas en Kazajistán. No hay una autoridad que lo haga cumplir, porque no es una norma. Lo que una persona puede hacer con él, por ahora, es entender qué se va a medir y qué no.

### EL CASO QUE RESPONDE

Esta semana no se encontró ningún caso real que ilustre lo que el protocolo quiere medir.

### LO QUE SE MIDIÓ

No se midió nada todavía. El artículo es un protocolo de estudio: describe qué se medirá, con qué escenarios y con qué reglas, pero no reporta ningún resultado, ninguna tasa de falla, ningún porcentaje de adecuación ni ninguna diferencia entre idiomas. No hay efecto que reportar porque no hay observaciones.

Lo que el protocolo sí especifica es el tamaño de la muestra que pretende tomar: 30 escenarios por módulo, 60 formulaciones emparejadas por módulo, 240 administraciones de primer intento en los dos idiomas obligatorios. Y advierte que ese conjunto no es una muestra representativa de las preguntas que hacen los ciudadanos: es un conjunto deliberadamente cargado hacia casos de riesgo. Si mañana se publicaran tasas de falla, no describirían lo que pregunta la mayoría de la gente.

### LA RESPUESTA DE OTRA REGIÓN AL MISMO ASUNTO

No se encontró esta semana ninguna corte, ley o autoridad de otra región que se pronuncie sobre el mismo método de auditar chatbots oficiales en varios idiomas.

### LA LECTURA DE LA BIBLIOTECA

Así lo lee weeklyAI. El protocolo trata las interfaces de inteligencia artificial enlazadas desde canales oficiales como una forma de control de la información pública: mira si sus respuestas coinciden con una referencia legal bloqueada, si conservan una guía materialmente equivalente en kazajo y en ruso, y si muestran rutas de verificación, corrección, revisión humana, queja y reparación.

Lo que un público puede comprobar solo, con lo que ve antes de votar, es limitado. Puede notar si una respuesta suena fluida. No puede notar si le omitieron un plazo, un requisito o la oficina competente, porque la omisión no deja huella visible. Para eso necesita que alguien más revise: que haya una fuente oficial contra la cual comparar, y que alguien con autoridad responda cuando la comparación falla.

Esta lectura se equivocaría si el estudio encuentra que las interfaces muestran rutas claras de corrección y revisión humana en ambos idiomas, y que esas rutas llevan a un responsable identificable con capacidad de responder, no solo a una página genérica de ayuda. También se equivocaría si las respuestas en ambos idiomas conservan por igual los datos decisivos y las omisiones son raras y detectables por el propio usuario sin comparar idiomas.

Lo que weeklyAI espera encontrar, dicho como expectativa y no como hallazgo: es probable que un chatbot oficial de información electoral suene creíble aunque omita un dato crítico, y que la misma pregunta reciba respuestas distintas según el idioma. También es probable que las rutas de corrección existan en el papel pero no estén conectadas a nadie que responda, sobre todo en el idioma con menos recursos técnicos.

Usted puede hacer algo con esto. Cuando use el chatbot de su autoridad electoral, pídale la fuente oficial y compárela con la resolución o el reglamento publicado. Si puede, haga la misma pregunta en los dos idiomas que ofrece y fíjese si le dan el mismo plazo y la misma oficina. Y pida una corrección o una revisión humana: vea si llega a una persona o solo a un formulario vacío.

### LO QUE NO SE SABE

No se sabe si el chatbot del gobierno kazajo o Sailau AI dan respuestas adecuadas, inadecuadas o desiguales en kazajo y en ruso. El protocolo no reporta ninguna salida, ninguna respuesta observada, ningún resultado.

No se sabe si los hallazgos, cuando existan, se aplicarían al resto del sector público de Kazajistán o a otros países. El propio protocolo advierte que su conjunto de desafíos no es una muestra de probabilidad y que las proporciones de falla describirían un corpus cargado de riesgo, no todas las preguntas de los ciudadanos.

No se sabe si los votantes notan una omisión, confían en la respuesta, cambian de comportamiento, abandonan un servicio o consiguen una corrección. El protocolo evalúa condiciones informativas, no la experiencia ciudadana2.

No se sabe qué modelo, qué datos de entrenamiento, qué configuración de recuperación o qué intención institucional hay detrás de cada respuesta. El diseño de caja negra observa lo que la interfaz muestra, no lo que ocurre adentro2.

No se sabe si las interfaces son legalmente vinculantes. El protocolo distingue entre lo que la interfaz muestra y lo que la ley obliga: la presentación oficial no confiere autoridad legal a una respuesta.

Lo que sí puede hacer usted, hoy, con lo que este protocolo le enseña: la próxima vez que su autoridad electoral le ofrezca un chatbot para consultar dónde votar o cómo impugnar, haga la misma pregunta en todos los idiomas que ofrezca y compare si le dan el mismo plazo y la misma oficina. Si las respuestas no coinciden, ya sabe algo que el chatbot no le dijo.

De dónde sale cada dato de contexto, y cuánto leímos de cada documento

  1. Tilep, B., Sultan, Y., Akaidar, K. et al. (2026). Auditing multilingual public information through officially linked AI interfaces in Kazakhstan. Frontiers in Political Science. https://doi.org/10.3389/fpos.2026.1984220 - el artículo del que trata esta nota — el artículo completo — el pasaje: "This protocol describes a prospective black-box audit of an e-government AI interface subject to prespecified identity checks and Sailau AI in Kazakhstan. It asks whether first-attempt outputs align with a locked legal-administrative reference, provide materially equivalent guidance in Kazakh and Russian, and display pathways for verification, correction, human review, complaint, and redress."
  2. Tilep, B., Sultan, Y., Akaidar, K. et al. (2026). Auditing multilingual public information through officially linked AI interfaces in Kazakhstan. Frontiers in Political Science. https://doi.org/10.3389/fpos.2026.1984220 - el artículo del que trata esta nota — el artículo completo — el pasaje: "A black-box design is appropriate because the training data, retrieval configuration, system prompt, model weights, and allocation of institutional responsibility are not publicly observable. The study can evaluate what a situated interface visibly returns under controlled inputs, but it cannot identify developer intent, hidden mechanisms, or effects on citizens"

Tilep, B., Sultan, Y., Akaidar, K. et al. (2026). Auditing multilingual public information through officially linked AI interfaces in Kazakhstan. Frontiers in Political Science. https://doi.org/10.3389/fpos.2026.1984220

Quién pagó: Los autores declararon que no se recibió apoyo financiero para este trabajo ni para su publicación, y declararon no tener relaciones comerciales o financieras que pudieran constituir un conflicto de interés.

other · Frontiers in Political Science · the paper, 5 Nov 2026 · free

A team in Kazakhstan has published a plan to test two official chatbots. It has not tested them yet.

Short version · the longer version follows, about 6 min

Ask the weeklyAI chat

Ask me about AI in your life: I point you to the stories weeklyAI has published on it, this week and before.

Conversations are saved for as long as weeklyAI exists, to improve the publication. Answers come in the language you write in.

The study at a glance
Who
Two official chatbots: an e-government AI interface and Sailau AI
How many
30 questions per bot
Where
Kazakhstan
When
Published 5 November 2026
Kind of study
review of studies
Who did it
Abai Kazakh National Pedagogical University and Asfendiyarov Kazakh National Medical University, Almaty
The limit that matters
A plan only: no answers have been asked for or scored, so it reports no results
The plan was published on 5 November 2026. It reports no answers and no results.
weeklyAI's reading

The plan comes from researchers at two universities in Almaty. It describes how they will question an e-government chatbot and Sailau AI, a bot the Central Election Commission links from its website through Telegram. Each bot gets its own set of 30 questions, in Kazakh and in Russian. The e-government bot is asked about everyday public services, and Sailau AI is asked about elections after the vote. The answers are compared against official rules. The plan was published on 5 November 2026. It reports no answers and no results.

The plan says it will record the ways each bot shows for checking an answer, asking for a correction, reaching a person, or making a complaint. It says it will note these in Kazakh and in Russian, and that seeing them does not prove they work.

The plan says both bots offer Kazakh and Russian. It says Sailau AI draws on election-law sources and an AlemLLM basis, as the commission reports. The authors write that presenting an answer through an official channel does not make it legally binding. The plan also says the commission held the election on 23 August 2026 and registered 145 deputies on 27 August.

What this means for you

For now there is nothing to check against your own election authority, because this team has published only a plan. What you can watch for is whether an official chatbot in your own country ever gets tested this way: questions in each language it offers, answers compared with the written rules, and the routes for correction written down. Until then, treat a fluent answer from any official bot as a claim to verify.

Tilep, B., Sultan, Y., Akaidar, K. et al. (2026). Auditing multilingual public information through officially linked AI interfaces in Kazakhstan. Frontiers in Political Science. https://doi.org/10.3389/fpos.2026.1984220

Who paid: The authors declared that no financial support was received for this work or its publication, and they declared no commercial or financial relationships that could be a conflict of interest.

The findings of other studies mentioned here come from this document; those studies were not opened one by one.

The longer versionThe passages copied from the paper, the pictures, and every source with how much of it we read · 1126 words · about 6 minRead it →Close

The rule and the evidence: A team in Kazakhstan has published a plan to test two official chatbots

### THE ACT

What is in force here is not a rule but a plan for a test, and it is worth being clear about that before anything else: the document published on 5 November 2026 is a study protocol, a prospective black-box audit of two artificial intelligence interfaces linked from official public-sector channels in Kazakhstan1. The two are an e-government AI interface subject to prespecified identity checks and Sailau AI in Kazakhstan1. The protocol asks three things: whether first answers match a locked legal-administrative reference, whether they give materially equivalent guidance in Kazakh and Russian, and whether they display pathways for verification, correction, human review, complaint, and redress1.

The plan calls for 240 core first-wave administrations in Kazakh and Russian, 120 per interface, and 336 core observations across repeat waves, rising to a maximum of 420 if a conditional English branch is activated. No output has been observed. The protocol reports no results, and the authors state that the original submission reported no study-eligible administration or output access1. A person reading this document cannot learn what either bot told anyone, because nobody has yet asked it anything that counts.

The legal setting is set out as a scope condition. Under Article 94(1), the 2026 Constitution entered into force on 1 July 2026; Article 9 establishes Kazakh as the state language and provides for the official use of Russian alongside Kazakh in state organizations and local self-government bodies2. The audit treats parity as material equivalence, not identical wording and not equal constitutional status2.

### THE CASE IT ANSWERS

No real instance was found this week: no named deepfake, avatar, bot network or manipulated tally form, with a place, a date and a second source, was located for this story.

### WHAT WAS MEASURED

Nothing has been measured yet. The article is a study protocol, and no passage in it gives a result, an adequacy rate or a disparity value1.

What the protocol specifies instead is a set of planned outcomes: requested-language material adequacy, directional Kazakh-Russian adequacy disparity, bilingual joint adequacy, and, among returned and legally evaluable outputs, critical and overall material information failure1. The sample is a deliberately risk-enriched challenge set of 30 scenarios per interface, not a probability sample of citizen questions1. The country is Kazakhstan, and the measurement has not been made in an election or anywhere else.

### THE OTHER REGION'S ANSWER TO THE SAME THING

None was found this week: no court, law or authority in another country, dated, on auditing an election authority's chatbot across its offered languages, appears in the passages read.

### THE LIBRARY'S READING

Here is weeklyAI's reading. A calm, official-sounding answer does part of the persuading by itself: people treat a tool as more credible than it has earned when it sits on an official site and speaks evenly, and the people running it have an interest in keeping that neutral image intact. weeklyAI expects that readers will extend trust to a voting-information chatbot for exactly that reason, and that an authority will be slow to publish anything showing its answers differ by language. This reading would be wrong if an election office posted its own side-by-side comparison of the languages it offers, or if people shown the same flawed answer in both languages rated it no more trustworthy than a plainly labelled draft.

A second reading. What counts as an official fact is made, not found: someone chose which rule, date and exception the bot repeats, and that choice is invisible to the person asking. weeklyAI expects a missing exception to read like a settled rule. This reading would be wrong if the answers consistently matched the full official text, exceptions and deadlines included, and the authority could show a documented procedure for how each rule was selected and updated.

A third, for households that speak two languages. When one group is cut off from information flowing to another, each settles into its own shared version of things, and the gap widens even while each version feels consistent inside its own channel. weeklyAI expects that if the Kazakh and Russian channels of the same bot draw on different material, users of each will settle into different understandings of the same election rule. This reading would be wrong if matched questions returned the same decision-critical content in both languages. What the reader can do with it: if you speak both, ask the same voting question in each language and compare the answers with a neighbour who uses the other. Report any gap to the election authority and to a local reporter.

### WHAT IS NOT KNOWN

The protocol is explicit about what a black-box design cannot see. The training data, retrieval configuration, system prompt, model weights, and allocation of institutional responsibility are not publicly observable; the study can evaluate what an interface visibly returns under controlled inputs, but it cannot identify developer intent, hidden mechanisms, or effects on citizens3. For Sailau AI, the Central Election Commission reports an AlemLLM basis, but the cited sources do not identify the deployed version, retrieval configuration, system prompt, hosting, or data roles, and the protocol does not infer them4.

Three cautions in the protocol are worth carrying to any future result. Completeness does not guarantee decision-critical fidelity; cross-language agreement does not establish parity when both outputs are materially inadequate; and source traceability does not by itself establish visible contestability5. Differences between languages may become politically significant if users rely on an answer that omits an eligibility condition, deadline, competent authority, or review route6. None of this has been observed, because no answer has been scored.

The study's architecture, the authors say, may be adapted to other multilingual public-sector settings only when case-specific legal authority, language status, platform governance, and data-protection requirements are newly established7. That is a boundary, not a promise: a finding in Kazakhstan would not travel to another country's election office on its own.

The protocol builds a chain of four links, a locked official reference standard, what the interface shows, what action it guides toward, and what routes for challenge are visible, and it treats each link as separately checkable8. That separation is the useful part for a reader anywhere: it gives you a list of things to look for, not a verdict.

What the record is silent on: whether anyone notices an omission, trusts an answer, changes a plan or gets a correction. The study measures the conditions of information, not what a voter does with them. So the question a voter can carry to the screen is this: when an official chatbot answers you, ask who wrote the rule it is repeating, and where the official text is.

Where each piece of context comes from, and how much of it we read

  1. Tilep, B., Sultan, Y., Akaidar, K. et al. (2026). Auditing multilingual public information through officially linked AI interfaces in Kazakhstan. Frontiers in Political Science. https://doi.org/10.3389/fpos.2026.1984220 - the article this story is about — the whole article — the passage: "This protocol describes a prospective black-box audit of an e-government AI interface subject to prespecified identity checks and Sailau AI in Kazakhstan. It asks whether first-attempt outputs align with a locked legal-administrative reference, provide materially equivalent guidance in Kazakh and Russian, and display pathways for verification, correction, human review, complaint, and redress."
  2. Tilep, B., Sultan, Y., Akaidar, K. et al. (2026). Auditing multilingual public information through officially linked AI interfaces in Kazakhstan. Frontiers in Political Science. https://doi.org/10.3389/fpos.2026.1984220 - the article this story is about — the whole article — the passage: "Kazakhstan is analytically important because language status and technical resources are not symmetrical. Under Article 94(1), the 2026 Constitution entered into force on 1 July 2026. Article 9 establishes Kazakh as the state language and provides for the official use of Russian alongside Kazakh in state organizations and local selfgovernment bodies"
  3. Tilep, B., Sultan, Y., Akaidar, K. et al. (2026). Auditing multilingual public information through officially linked AI interfaces in Kazakhstan. Frontiers in Political Science. https://doi.org/10.3389/fpos.2026.1984220 - the article this story is about — the whole article — the passage: "A black-box design is appropriate because the training data, retrieval configuration, system prompt, model weights, and allocation of institutional responsibility are not publicly observable. The study can evaluate what a situated interface visibly returns under controlled inputs, but it cannot identify developer intent, hidden mechanisms, or effects on citizens"
  4. Tilep, B., Sultan, Y., Akaidar, K. et al. (2026). Auditing multilingual public information through officially linked AI interfaces in Kazakhstan. Frontiers in Political Science. https://doi.org/10.3389/fpos.2026.1984220 - the article this story is about — the whole article — the passage: "Sailau AI is accessed from the CEC website through @SailauAIbot on Telegram (Central Election Commission of the Republic of Kazakhstan, n.d.). The CEC reports an AlemLLM basis, but cited sources do not identify the deployed version, retrieval configuration, system prompt, hosting, or data roles; these are not inferred."
  5. Tilep, B., Sultan, Y., Akaidar, K. et al. (2026). Auditing multilingual public information through officially linked AI interfaces in Kazakhstan. Frontiers in Political Science. https://doi.org/10.3389/fpos.2026.1984220 - the article this story is about — the whole article — the passage: "First, completeness does not guarantee decision-critical fidelity. Second, cross-language agreement does not establish parity when both outputs are materially inadequate. Third, source traceability does not by itself establish visible contestability."
  6. Tilep, B., Sultan, Y., Akaidar, K. et al. (2026). Auditing multilingual public information through officially linked AI interfaces in Kazakhstan. Frontiers in Political Science. https://doi.org/10.3389/fpos.2026.1984220 - the article this story is about — the whole article — the passage: "Such differences may become politically significant if users rely on an answer that omits an eligibility condition, deadline, competent authority, or review route."
  7. Tilep, B., Sultan, Y., Akaidar, K. et al. (2026). Auditing multilingual public information through officially linked AI interfaces in Kazakhstan. Frontiers in Political Science. https://doi.org/10.3389/fpos.2026.1984220 - the article this story is about — the whole article — the passage: "The architecture may be adapted to other multilingual public-sector settings only when case-specific legal authority, language status, platform governance, and data-protection requirements are newly established."
  8. Tilep, B., Sultan, Y., Akaidar, K. et al. (2026). Auditing multilingual public information through officially linked AI interfaces in Kazakhstan. Frontiers in Political Science. https://doi.org/10.3389/fpos.2026.1984220 - the article this story is about — the whole article — the passage: "The study evaluates a four-part normative chain: a locked official reference standard, observable output representation, observable action guidance, and visible routes for verification, correction, human review, complaint, and redress."

Tilep, B., Sultan, Y., Akaidar, K. et al. (2026). Auditing multilingual public information through officially linked AI interfaces in Kazakhstan. Frontiers in Political Science. https://doi.org/10.3389/fpos.2026.1984220

Who paid: The authors declared that no financial support was received for this work or its publication, and they declared no commercial or financial relationships that could be a conflict of interest.

The free weekly drip by email

One email a week: the free story of every section, with its picture, and the headlines of the rest.

No payment, nothing to cancel. Unsubscribe from any issue.

El goteo semanal gratis por correo

Un correo por semana: la nota gratuita de cada sección, con su imagen, y los titulares de las demás.

Sin pago y sin nada que cancelar. Puede darse de baja desde cualquier envío.

From the house

About

Every week the new research on AI is read here, and this page tells you what was found, how far it can be trusted, and what you can do with it. This page is free, and every line on it says what was found.

The studies are read by a machine built and ruled by a professor of communication at Sacramento State and Universidad Nacional Autónoma de México (UNAM). Who makes this →

Diego Bonilla, Ph.D.

Full Professor of Communication Studies, Sacramento State, California, United States
Graduate instructor, Universidad Nacional Autónoma de México (UNAM), Mexico

hypergraphia.com

I am a professor of communication at Sacramento State, where I created the course Information Management and Privacy and have taught it many times, and at UNAM, in Mexico City, I teach the graduate course Electronic and Generative Literature with AI. In both I teach AI. I have taught since 1989, at first mathematics and physics to children who found them hard, and since 1997 one subject that has changed its name many times: media literacy, computer literacy, digital literacy, information literacy, ICT literacy, and now AI literacy.

This page exists because most people cannot go back to a university, and these systems are changing their work, their children's schooling, their parents' care and their vote all the same. What a university teaches about them should reach the people who need it, every week, in plain language, without enrolling anywhere. That is what I do here, and it is what I do on Thursday mornings in a seminar for the Renaissance Society at Sacramento State, built on this publication, where anyone can send a question and hear it answered.

In 2003 my doctorate at Syracuse measured what a machine can infer about a person from the traces they leave, and what I found became the privacy course I still teach. In 2007 and 2009 I sat on the committee at ETS in Princeton that set the national standard for digital literacy, and from 2014 to 2016 I was the California State University's ICT Literacy Ambassador. Six names, one job: helping people not be at the mercy of the newest machine.

I have also worked on the other side of the counter. In 2000 I began making films that a machine assembles differently for every viewer; the first of them took first prize at the Moscow International Film Festival in 2004, two decades before the industry shipped a generative model. A generative video poem I made with a poet at UNAM is in the field's definitive anthology, and in 2025 my AI short film The Human Futures Market took Best AI Short Film at the Red Rocks AI Film Festival. My recent published research is on AI and elections, and on how these models represent the cultures they were not trained on. I do not read the papers myself, and this page will never tell you I do. I built the machine that reads them and wrote the rules it reads by, and I analyze what it gives me the way I have taught my students to analyze: asking how it knows, and who benefits from what it says.

The vendors began teaching AI literacy two years ago, because before that there was nothing to sell. I am not against them. I am simply not one of them, and no one who sells these systems pays for this page. Your subscription does, and that is what keeps it honest.

The record, dated
  • 1989 Teaching begins, Mexico City
  • 1995 Hypergraphia, an early Spanish-language web and hypermedia publisher
  • 1996 Mercadotecnia e imagen en Internet, with Jesús del Toro, a book on marketing on the Internet, among the first in Spanish, Editorial Iberoamérica
  • 1999 Autorretrato, first generative poem, published by UNAM's Instituto de Investigaciones Filológicas
  • 2003 Ph.D., Syracuse University, Doctoral Prize, for the dissertation The Medium Is the Measure of Itself
  • 2004 A Space of Time: First Prize, Internet/Multimedia, XXVI Moscow International Film Festival
  • 2007, 2009 ETS iSkills committee, Princeton
  • 2008 Making Sense of Tracking Data, a book on how the traces a person leaves online allow inferences about them, including how they think, VDM Verlag
  • 2011 Keynote on generative work, UNAM
  • 2014–2016 ICT Literacy Ambassador, California State University · California Open Educational Resources Council
  • 2019 Big Data, with Rodolfo Mata: Electronic Literature Collection, Volume 4
  • 2020 Tablada Hipertextual, with Rodolfo Mata, UNAM: the computer reads a poet's words and builds a different reading from them, the ground the language models now work on
  • 2021 Uku Pacha, a generative VR film: world premiere, Vancouver International Film Festival
  • 2023 Keynote on the representation of Latin American cultures in AI, Ibero-American Meeting of Audiovisual Narratives, Trujillo, Peru · Graduate teaching at UNAM begins
  • 2023–2026 Directed the faculty learning community Advancing AI Literacy, Sacramento State · Principles for Cognitive and Pedagogical Sovereignty
  • 2024 A weekly AI publication run on real pipelines, every week since
  • 2025 Best AI Short Film, Red Rocks AI Film Festival
  • 2026 Keynote, with Nancy Huante-Tzintzun, Latin American and Caribbean Digital Humanities symposium, UT San Antonio · AI Watch, a weekly seminar for the Renaissance Society at Sacramento State

The research record →

TertulIA: weeklyAI.watch/tertulia/en/Pictures are illustrations generated by weeklyAI.watch, never photographs