Weekly AI Watch

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

Week of 13 November 2026 | Past editions15,811 papers found · 14,944 read · 4,785 belonged · 46 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 13 de noviembre de 2026 | Ediciones anteriores15,811 estudios encontrados · 14,944 leídos · 4,785 correspondían · 46 notas pasaron todas las comprobaciones

weeklyAI · Week of 13 November 2026weeklyAI · Semana del 13 de noviembre de 2026

← Your vote← Su voto

survey · Nepal Journal of Multidisciplinary Research · la publicación, 1 jul 2026 · gratis

Lo que un sondeo entre 400 usuarios de redes en Nepal dice sobre los videos falsos, y lo que no dice

Versión breve · la versión detallada sigue, unos 7 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
Personas en Nepal que usan redes y medios digitales varias veces por semana
Cuántos
400
Dónde
Nepal
Cuándo
Entre septiembre de 2025 y enero de 2026
Tipo de estudio
survey
Quién lo hizo
University of Wolverhampton y Herald College Kathmandu
El límite que importa
Mide lo que la gente opina, no si de verdad detecta un video falso; solo en Nepal
Es una encuesta de percepción, no una prueba de detección: a nadie se le puso frente a un video falso para ver si lo identificaba.
Lectura de weeklyAI

1. Entre septiembre de 2025 y enero de 2026, una investigadora preguntó a 400 personas en Nepal que usan redes y medios digitales varias veces por semana cómo calificaban los esfuerzos del país contra los videos falsos y las noticias inventadas. Siete de cada diez tenían entre 20 y 30 años. Más de tres cuartas partes habían terminado una carrera. Quienes mejor calificaron esos esfuerzos fueron los que decían haber recibido campañas de concientización y tener más habilidades para revisar lo que ven. Los sistemas de seguridad, las leyes y las medidas de las plataformas no mostraron una relación clara con esa calificación. El estudio explica solo el 35.3 por ciento de las respuestas.

2. El estudio no propone ninguna forma de que usted revise esto por su cuenta, y no cubre registros, conteos ni elecciones de América. Para un video sospechoso de un candidato, busque la versión original en el canal oficial de esa persona y compare fecha, voz y escena.

3. No hay autoridad electoral citada en el documento.

Qué significa para usted

Para usted, que va a votar, el sondeo no ofrece ninguna garantía sobre su registro, su casilla o el conteo: se hizo en Nepal, sin autoridad electoral citada. Lo que sí puede hacer es revisar quién firma el material que recibe y buscar la versión original en el canal oficial del candidato. Tampoco dé por hecho que las leyes o las plataformas lo están resolviendo.

Kumari, A. (2026). AI and Cybersecurity: Combating Deepfakes and Misinformation in Nepal. Nepal Journal of Multidisciplinary Research. https://doi.org/10.3126/njmr.v9i2.95265

Quién pagó: El artículo no indica quién financió el estudio, si algún financiador tuvo voz en él, ni quién prestó equipos o programas informáticos.

Versión detalladaLos pasajes copiados del artículo, las ilustraciones y cada fuente con cuánto leímos de ella · 1341 palabras · unos 7 minLeerla →Cerrar

La regla y la evidencia: lo que un sondeo entre 400 usuarios de redes en Nepal dice sobre los videos falsos, y lo que no dice

### EL ACTO, CITADO Y ENLAZADO

Esta semana no hay un acto que citar. El documento que se revisó no es una resolución, ni una sentencia, ni una auditoría, ni un proyecto de ley: es un artículo académico sobre percepciones, publicado en junio de 2026 en una revista nepalí1. Nadie queda obligado por él. No dice quién lo aplica, ni desde cuándo, ni qué puede hacer una persona y para cuándo.

Lo más cercano a una regla está en sus recomendaciones, y está escrito como consejo, no como norma: la educación en medios debería integrarse a los planes de estudio, y las campañas deberían enfocarse en verificar contenido, pensar con criterio y comportarse con responsabilidad en línea2. También pide al gobierno fortalecer sus políticas de ciberseguridad y mejorar la transparencia de la moderación en las plataformas3. Nada de eso está en vigor por este texto.

El estudio sí describe el terreno donde una regla tendría que operar. Los sistemas que generan video falso funcionan con dos redes neuronales que compiten entre sí para que el material salga cada vez más realista4. Con esas herramientas se pueden fabricar discursos y hechos atribuidos a figuras públicas5. Y los sistemas que recomiendan contenido en las redes priorizan lo que genera reacción, aunque sea falso, de modo que la mentira llega a mucha gente antes de que alguien la revise6.

### EL CASO QUE RESPONDE

Esta semana no se encontró ningún caso real que contar.

### LO QUE SE MIDIÓ

Aquí hay una medición, y conviene mirarla de cerca antes de sacar conclusiones. Se encuestó a 400 personas en Nepal que usan redes y medios digitales con regularidad, de entre menos de 20 y más de 50 años, elegidas a propósito. Se les preguntó, en una escala de cinco puntos, qué tan eficaz les parecía la lucha contra los videos falsos y la desinformación en su país.

El resultado principal: las campañas de concientización pública y la alfabetización digital se asociaron con una mayor eficacia percibida7. En el modelo, las campañas obtuvieron un coeficiente de 0.476 y la alfabetización digital uno de 0.1778. El modelo completo explicó el 35.3% de la variación en las respuestas. Es decir, casi dos tercios de lo que la gente respondió no se explica por ninguna de las cinco cosas que se midieron9.

Las otras tres no salieron bien libradas. La infraestructura de ciberseguridad y las políticas de gobierno no mostraron efecto significativo8. Las medidas de las plataformas sociales salieron con signo negativo10. Los autores también señalan que ese signo negativo no prueba una ineficacia real y que puede reflejar la percepción de los encuestados sobre la moderación11. Su diseño no permite profundizar en esa percepción, y lo reconocen12.

Hay que decir con claridad qué es y qué no es esto. Es una encuesta de percepción, no una prueba de detección: a nadie se le puso frente a un video falso para ver si lo identificaba. Mide lo que la gente cree, en Nepal, entre usuarios habituales de redes. No se ha medido en una elección, ni en un laboratorio con tareas de detección, ni en ningún país de América.

Otros trabajos, en cambio, sí pusieron a prueba la vista de la gente. En un experimento con imágenes generadas por inteligencia artificial, las personas no fueron sustancialmente mejores que el azar para distinguir las falsas de las auténticas13. Solo los menores de 25 años mostraron un desempeño claramente superior al azar14. Y los propios autores advierten que las cosas pueden empeorar conforme la tecnología mejore, porque la gente ya batalla incluso con imágenes de calidad baja15. Otro trabajo reporta que los evaluadores humanos reconocen las falsificaciones de alta calidad apenas una cuarta parte de las veces16.

### LA RESPUESTA DE LA OTRA REGIÓN A LO MISMO

Esta semana no encontramos ninguna corte, ley ni autoridad de otro país que se haya pronunciado sobre este método.

### LA LECTURA DE LA BIBLIOTECA

Así lo lee weeklyAI. El rumor falso sobre un candidato no es nuevo: siempre ha trabajado con material auténtico sacado de contexto, con fotos retocadas y con datos fabricados, y casi siempre llega por alguien de confianza. Lo que cambia ahora es la velocidad y el acabado: lo que antes había que imaginar, hoy se puede ver y oír.

De ahí sale una expectativa que weeklyAI pone sobre la mesa, y la pone como expectativa, no como hallazgo. En una casa como la del lector, el video falso de campaña llegará por un grupo de mensajes, entre un saludo familiar y una broma, en el momento en que nadie está revisando nada. Es probable que se crea ahí, en ese descanso, y no en la discusión política donde uno ya va con la guardia levantada. Esta lectura se equivocaría si resultara que quien ve política por puro gusto o por costumbre distingue los videos falsos igual o mejor que quien los revisa desconfiando a propósito.

Hay un costo que conviene nombrar sin adornos. Revisar cansa. Consultar la resolución, la auditoría o el informe de la autoridad electoral cuesta más que leer el meme que llegó primero, y por eso la mayoría se queda con el resumen. No es flojera: es que la atención se acaba. Lo que se gana con el atajo es rapidez; lo que se entrega es la posibilidad de saber si el conteo que se está denunciando en un video tiene algún documento detrás.

Y hay un hueco que el descrédito institucional agranda. Cuando la autoridad electoral o el tribunal llegan desgastados a la jornada, el video que denuncia un fraude empieza a circular como si fuera prueba suficiente, y la gente le pide explicaciones al video antes que a la auditoría. Esta lectura se equivocaría si, aun con la autoridad desprestigiada, los votantes esperaran el informe oficial antes de dar por cierto un fraude mostrado en pantalla.

¿Qué puede hacer el lector con esto? Una sola cosa, pequeña y concreta: cuando llegue un video que denuncia algo del conteo, no lo desmienta ni lo comparta de inmediato; pregunte qué documento oficial lo respalda. Si es trabajador electoral, maestro o reportero local, diga en voz alta qué está verificado y qué es solo una grabación. Esa frase, dicha a tiempo, vale más que cualquier campaña.

### LO QUE NO SE SABE

No se sabe si las campañas de concientización realmente ayudan a alguien a detectar un video falso, porque el estudio mide lo que la gente opina, no lo que logra identificar. No se sabe si estos resultados valdrían fuera de Nepal, ni siquiera entre nepalíes que no usan redes con regularidad: la muestra se eligió a propósito entre usuarios habituales, muchos de ellos jóvenes y con estudios universitarios.

No se sabe por qué las medidas de las plataformas salieron con signo negativo; los autores lo dejan como pregunta abierta. No se sabe si la desconfianza hacia las plataformas precede a la mala calificación o si la mala calificación produce la desconfianza, porque el estudio se tomó en un solo momento. Y no se sabe nada, por este documento, sobre el registro electoral, la campaña, la votación o el conteo en América Latina, Estados Unidos o Canadá: la encuesta es de Nepal y ahí se queda.

Tampoco se sabe cuánto aguantan los detectores automáticos en condiciones reales. Un trabajo reporta que los métodos habituales fallan con la compresión que aplican las redes sociales y que sus puntajes no se pueden explicar ante un juez17. Otro más, que las falsificaciones también sirven para fraude financiero y suplantación18.

Esta lectura se equivocaría si el lector encontrara que en su propia casa la gente ya consulta la resolución o la auditoría antes de creer una denuncia en video. Ojalá.

Antes de votar, cuando le llegue un video que le dice cómo va a estar el conteo, pregúntese quién lo grabó y qué papel oficial lo respalda, y espere ese papel antes de decidir.

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

  1. Kumari, A. (2026). AI and Cybersecurity: Combating Deepfakes and Misinformation in Nepal. Nepal Journal of Multidisciplinary Research. https://doi.org/10.3126/njmr.v9i2.95265 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Deepfake technology represents one of the most disruptive applications of AI, enabling the creation of highly realistic synthetic audio-visual content. These systems rely on deep learning models that can replicate human facial expressions, voice patterns, and behaviors with high accuracy (Mirsky & Lee, 2021)."
  2. Kumari, A. (2026). AI and Cybersecurity: Combating Deepfakes and Misinformation in Nepal. Nepal Journal of Multidisciplinary Research. https://doi.org/10.3126/njmr.v9i2.95265 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Media literacy education should be integrated into academic curricula, while awareness campaigns should focus on content verification, critical thinking, and responsible online behavior."
  3. Kumari, A. (2026). AI and Cybersecurity: Combating Deepfakes and Misinformation in Nepal. Nepal Journal of Multidisciplinary Research. https://doi.org/10.3126/njmr.v9i2.95265 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Additionally, the government should strengthen AI-related cybersecurity policies, encourage public-private collaboration for deepfake detection technologies, and improve transparency in platform-based moderation systems."
  4. Kumari, A. (2026). AI and Cybersecurity: Combating Deepfakes and Misinformation in Nepal. Nepal Journal of Multidisciplinary Research. https://doi.org/10.3126/njmr.v9i2.95265 - el artículo del que trata esta nota — el artículo completo — el pasaje: "The advancement of generative adversarial networks (GANs) has further accelerated the development of deepfake technologies. These models operate through adversarial training, where two neural networks compete to improve the realism of generated content (Tolosana et al., 2021)."
  5. Kumari, A. (2026). AI and Cybersecurity: Combating Deepfakes and Misinformation in Nepal. Nepal Journal of Multidisciplinary Research. https://doi.org/10.3126/njmr.v9i2.95265 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Additionally, deepfakes can influence political discourse by fabricating speeches and events involving public figures (Bennett & Livingston, 2021)."
  6. Kumari, A. (2026). AI and Cybersecurity: Combating Deepfakes and Misinformation in Nepal. Nepal Journal of Multidisciplinary Research. https://doi.org/10.3126/njmr.v9i2.95265 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Algorithm-driven recommendation systems prioritize engagement-based metrics, often amplifying emotionally charged or sensational content regardless of its factual accuracy. This structural design enables misleading information to spread rapidly across interconnected user networks, often reaching large audiences before verification mechanisms can intervene (Cinelli et al., 2021; Zannettou et al., 2022)."
  7. Kumari, A. (2026). AI and Cybersecurity: Combating Deepfakes and Misinformation in Nepal. Nepal Journal of Multidisciplinary Research. https://doi.org/10.3126/njmr.v9i2.95265 - el artículo del que trata esta nota — el artículo completo — el pasaje: "The findings reveal that Public Awareness Campaigns and Digital Literacy significantly and positively influence the effectiveness of combating deepfakes and misinformation."
  8. Kumari, A. (2026). AI and Cybersecurity: Combating Deepfakes and Misinformation in Nepal. Nepal Journal of Multidisciplinary Research. https://doi.org/10.3126/njmr.v9i2.95265 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Regression analysis indicates that Public Awareness Campaigns (B = 0.476, p < 0.001) and Digital Literacy (B = 0.177, p = 0.008) are significant predictors, while Cybersecurity Infrastructure, Government Policies, and Social Media Platform Measures show non-significant or weak effects."
  9. Kumari, A. (2026). AI and Cybersecurity: Combating Deepfakes and Misinformation in Nepal. Nepal Journal of Multidisciplinary Research. https://doi.org/10.3126/njmr.v9i2.95265 - el artículo del que trata esta nota — el artículo completo — el pasaje: "This suggests that approximately 64.7% of the variation remains unexplained by the model. Therefore, other factors not examined in this study, such as trust in media, political orientation, social network influence, fact-checking behavior, technological experience, and demographic characteristics, may also contribute"
  10. Kumari, A. (2026). AI and Cybersecurity: Combating Deepfakes and Misinformation in Nepal. Nepal Journal of Multidisciplinary Research. https://doi.org/10.3126/njmr.v9i2.95265 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Specifically, SMPM has a notable negative effect (β = -0.126, p = 0.041), implying that existing actions taken by social media platforms may be ineffective or even harmful in reducing misinformation within Nepal's context."
  11. Kumari, A. (2026). AI and Cybersecurity: Combating Deepfakes and Misinformation in Nepal. Nepal Journal of Multidisciplinary Research. https://doi.org/10.3126/njmr.v9i2.95265 - el artículo del que trata esta nota — el artículo completo — el pasaje: "The negative coefficient may not necessarily indicate that platform measures are objectively ineffective. Instead, it may reflect respondents' perceptions regarding the adequacy, transparency, consistency, or trustworthiness of current platform moderation practices."
  12. Kumari, A. (2026). AI and Cybersecurity: Combating Deepfakes and Misinformation in Nepal. Nepal Journal of Multidisciplinary Research. https://doi.org/10.3126/njmr.v9i2.95265 - el artículo del que trata esta nota — el artículo completo — el pasaje: "A further limitation of this study is that the quantitative design does not allow for an in-depth exploration of respondent's perceptions regarding social media platform measures."
  13. Hove MF, Hjorth F, Golovchenko Y, Adler-Nissen R, Bechmann A, van Dalen A, et al. (2026). Seeing Is Believing? Most People Struggle to Detect AI-Generated Images. Social Media + Society. 10.1177/20563051261483376 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "In a preregistered survey experiment, we find that, on average, people are not substantially better than chance at discerning AI-generated images from authentic images, regardless of the context of the image."
  14. Hove MF, Hjorth F, Golovchenko Y, Adler-Nissen R, Bechmann A, van Dalen A, et al. (2026). Seeing Is Believing? Most People Struggle to Detect AI-Generated Images. Social Media + Society. 10.1177/20563051261483376 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "While younger people, highly analytical-thinking individuals, and experienced users of generative AI have better discernment abilities, no demographic group other than young individuals (<25) performs substantially better than chance."
  15. Hove MF, Hjorth F, Golovchenko Y, Adler-Nissen R, Bechmann A, van Dalen A, et al. (2026). Seeing Is Believing? Most People Struggle to Detect AI-Generated Images. Social Media + Society. 10.1177/20563051261483376 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "As generative AI technology likely advances and the realism of AI-generated images continues to improve, these findings are concerning, as people already struggle to discern between AI-generated and authentic images, even in an experiment where they have been primed and are assessing images of relatively low quality."
  16. Hussain M, Saeed F, Aldera S. (2026). Bioinspired Adaptive-Depth Neural Growth for Deepfake Video Forensics: An Entropy-Guided State-Space Framework. Biomimetics. 10.3390/biomimetics11090653 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "human evaluators accurately recognize high-quality forgeries merely 25% of the time, highlighting the pressing necessity for automated, widely applicable detection mechanisms"
  17. Deshpande S, Lajurkar S, Sarak P, Jain PPA. (2026). VeriMedia: A Spatial-Temporal Deep Learning Platform for Video Deepfake Detection and Forensic Verification. IJARCCE. https://doi.org/10.17148/ijarcce.2026.1511115 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "Along these dimensions, suffering from acute vulnerability to lossy social media compression (such as H.264/H.265 codecs at high quantization parameters c23 and c40); in fact, also black-box deep models output uninterpretable scalar scores without spatial or temporal attribution, restricting their evidentiary admissibility in judicial and regulatory environments."
  18. Deshpande S, Lajurkar S, Sarak P, Jain PPA. (2026). VeriMedia: A Spatial-Temporal Deep Learning Platform for Video Deepfake Detection and Forensic Verification. IJARCCE. https://doi.org/10.17148/ijarcce.2026.1511115 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "The increase of hyper-realistic generative models, deepfakes, face-swapping algorithms, and neural rendering has severely compromised public trust in digital video media, helping centered political disinformation, financial fraud, executive impersonation, and non-consensual synthetic defamation, in many respects"

Kumari, A. (2026). AI and Cybersecurity: Combating Deepfakes and Misinformation in Nepal. Nepal Journal of Multidisciplinary Research. https://doi.org/10.3126/njmr.v9i2.95265

Quién pagó: El artículo no indica quién financió el estudio, si algún financiador tuvo voz en él, ni quién prestó equipos o programas informáticos.

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

survey · Nepal Journal of Multidisciplinary Research · the paper, 1 Jul 2026 · free

Nepal survey of 400 finds awareness campaigns linked to people feeling better able to fight deepfakes

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
People in Nepal who use social media and online news several times a week
How many
400 respondents
Where
Nepal
When
September 3, 2025, to January 21, 2026
Kind of study
survey
Who did it
University of Wolverhampton and Herald College Kathmandu
The limit that matters
It asked people to rate the effort, not to spot a fake.

How the five answers to deepfakes scored in the Nepal survey

Public awareness campaignsagainstCybersecurity systems, government policies and platform moderation

Awareness campaigns scored highest and clearly positive; the other three were not significant or slightly negative

Digital literacyagainstCybersecurity systems, government policies and platform moderation

Digital literacy scored positive and significant; the other three were not significant or slightly negative

Platform moderation as respondents rated itagainstThe authors' caution about that rating

The negative score may reflect what people think of the moderation, not that it is useless

The 400 rated how well Nepal is fighting deepfakes. They were not tested on spotting one.
weeklyAI's reading

In a survey of 400 people in Nepal, run between September 2025 and January 2026, awareness campaigns were linked to better ratings of the country's fight against deepfakes, and cybersecurity systems, government policies and platform moderation were not. The author chose people who use social media and online news several times a week, and the people she aimed for ranged from under 20 to over 50. The author shared the questionnaire with about 500 people and 400 completed it. They chose people who use social media and online news, and most were aged 20 to 30 and had a university degree. The 400 rated how well Nepal is fighting deepfakes. They were not tested on spotting one.

The article names no register page, ad library or fact-checker a voter can consult, and it describes no way to check a video yourself.

It reports no rule or ruling from any electoral authority. It records only the survey's own result: campaigns scored highest (B = 0.476), digital literacy next (B = 0.177), and the model explained 35.3% of the variance.

What this means for you

For your own vote, the survey offers no checklist for spotting a faked video, and it cannot tell you whether awareness drives better detection or simply travels with it. What it does suggest is that in Nepal, people who use digital media rated public awareness and digital literacy above cybersecurity systems, government policies and platform moderation as ways to fight deepfakes. Watch whether your electoral authority publishes its audits and findings in plain language a voter can read.

Kumari, A. (2026). AI and Cybersecurity: Combating Deepfakes and Misinformation in Nepal. Nepal Journal of Multidisciplinary Research. https://doi.org/10.3126/njmr.v9i2.95265

Who paid: The article does not say who funded the study, whether any funder had a say, or who lent equipment or software.

The longer versionThe passages copied from the paper, the pictures, and every source with how much of it we read · 1239 words · about 6 minRead it →Close

The rule and the evidence: A Nepal survey of 400 digital media users found awareness campaigns and digital literacy were linked to a better rating of the effort, not to spotting a fake

### THE ACT, quoted exactly and linked

The document is not a rule, a ruling or an audit. It is a survey: 400 people in Nepal, reached online between September 3, 2025, and January 21, 2026, who answered a structured questionnaire about how well their country is fighting deepfakes and false news1. The journal received it on April 30, 2026 and accepted it on June 29, 2026, and it governs no one. No voter is bound by it, no authority enforces it, and there is no date by which a person must do anything.

What the study set out to test can be put plainly. It asked respondents to rate five possible answers to the problem: strong computer systems for catching fakes, a public that can read a screen critically, laws and regulations, action by the social media companies, and public awareness campaigns2. Each was scored on a five-point scale, and the scores were compared against how well respondents thought Nepal was doing at identifying and controlling deepfakes and misinformation2.

The limits belong beside the act from the first line. The outcome is a rating, not a test: no one was shown a fake video and asked to judge it. The sample was chosen on purpose, not drawn at random. And the model left most of the variation unexplained: about 64.7 percent of it3.

### THE CASE IT ANSWERS

No real instance was found this week: no named fake video, no named place, no date, no second source.

What can be said is what the study itself describes in general terms. The article describes a study in which deepfake tools can copy a face, a voice and a way of moving with high accuracy4. The article describes a study in which they are built by two programs set against each other, each one pushing the other toward more convincing output5. The article describes a study in which the result can put words in the mouth of a public figure, in a fabricated speech or a fabricated event6.

Then the platforms carry it. The article describes a study in which recommendation systems reward what holds attention, and what holds attention is often what stirs feeling, whether or not it is true, so a falsehood can reach a large audience before anyone checks it7. It would seem this is the kind of thing they mean.

### WHAT WAS MEASURED

Two factors stood out as positive and statistically significant: public awareness campaigns and digital literacy1. In the regression, awareness campaigns carried a coefficient of 0.476 with a p value below 0.001, and digital literacy carried 0.177 with a p value of 0.0082. Picture the awareness score as the heaviest hand on the scale, and the literacy score as a lighter one still clearly on the same side. The other three did not behave that way: computer systems and government policy were not significant, and platform measures came out slightly negative, at minus 0.126 with a p value of 0.04128. The whole model explained 35.3 percent of the variance2.

The negative platform figure needs care. The authors write that it may not mean the measures are objectively useless; it may reflect what respondents think of their adequacy, transparency, consistency or trustworthiness9. The study's own design cannot dig into that, which the authors concede10.

The sample was 400 respondents out of roughly 500 approached, an 80 percent response rate, aged from below 20 to 50 and above, mostly between 20 and 30 and mostly university-educated. It was measured in Nepal, in a survey, not in an election. No passage here reports a measurement taken at a polling station, a register or a count.

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

We found none this week.

### THE LIBRARY'S READING

Here is weeklyAI's reading. A false story about a candidate has always worked less by proving anything than by leaving a mark: once a charge is in the air, the accused carries the smell of it, and the person who repeats it can say they only asked11. What is new is the cost. A convincing face and voice used to take a studio; now they take an afternoon. What a public can learn to check is a source, a date, a second account of the same event. What it cannot check alone is a clip built to look exactly like the thing it is imitating, which is why the checking has to be done by more than one person, and out loud.

This reading leads weeklyAI to expect something specific in homes like the reader's. A voter who has been taught to ask who is speaking, what they gain and how the claim could be checked will do better than one who trusts a polished clip because it looks official. That is an expectation, not a result. This reading would be wrong if people drilled in those questions spotted fakes no better than people who were not. What the reader can do with it: before passing on a video or a claim about the register or the count, find a second source that does not simply repeat the first.

A second expectation follows from the same place. Warnings alone tend to slide off; practice in comparing sources tends to stay. So a voter who sees one viral clip and stops there is the one most exposed, and a poll worker or local reporter who checks a single official statement may miss a contradiction sitting in the next document. This reading would be wrong if people who consulted several sources were fooled as often as those who relied on one. What the reader can do with it: when a number about the count appears, ask where it came from and what is missing from it, and say so to the people around you, corrections included.

### WHAT IS NOT KNOWN

The study reports that awareness campaigns and literacy were significant predictors, but it does not say that either caused anyone to detect a fake12. The model left about 64.7 percent of the variation unexplained, and the authors say other factors not examined in the study may also play a part3.

Other work complicates the picture. The article describes a study in which, in a preregistered experiment, people on average were not substantially better than chance at telling AI-made images from real ones, and no group except the young did much better1213. The article describes another study in which human evaluators recognized high-quality forgeries only 25 percent of the time14. That is lab work, with images and videos served up for judgment, not a Nepali election.

The record is silent on several things a voter would want. It does not name a single deepfake seen in Nepal. It does not say what the respondents were shown, if anything. It does not report confidence intervals. And it does not touch a register, a campaign's targeting, a ballot or a count.

The last word belongs to the thing this piece opened on: a questionnaire filled in by 400 people about how well their country is doing. Before you vote, ask of any clip of a candidate or any claim about the count: who is asserting this, and how could I check it without them?

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

  1. Kumari, A. (2026). AI and Cybersecurity: Combating Deepfakes and Misinformation in Nepal. Nepal Journal of Multidisciplinary Research. https://doi.org/10.3126/njmr.v9i2.95265 - the article this story is about — the whole article — the passage: "The findings reveal that Public Awareness Campaigns and Digital Literacy significantly and positively influence the effectiveness of combating deepfakes and misinformation."
  2. Kumari, A. (2026). AI and Cybersecurity: Combating Deepfakes and Misinformation in Nepal. Nepal Journal of Multidisciplinary Research. https://doi.org/10.3126/njmr.v9i2.95265 - the article this story is about — the whole article — the passage: "Regression analysis indicates that Public Awareness Campaigns (B = 0.476, p < 0.001) and Digital Literacy (B = 0.177, p = 0.008) are significant predictors, while Cybersecurity Infrastructure, Government Policies, and Social Media Platform Measures show non-significant or weak effects."
  3. Kumari, A. (2026). AI and Cybersecurity: Combating Deepfakes and Misinformation in Nepal. Nepal Journal of Multidisciplinary Research. https://doi.org/10.3126/njmr.v9i2.95265 - the article this story is about — the whole article — the passage: "This suggests that approximately 64.7% of the variation remains unexplained by the model. Therefore, other factors not examined in this study, such as trust in media, political orientation, social network influence, fact-checking behavior, technological experience, and demographic characteristics, may also contribute"
  4. Kumari, A. (2026). AI and Cybersecurity: Combating Deepfakes and Misinformation in Nepal. Nepal Journal of Multidisciplinary Research. https://doi.org/10.3126/njmr.v9i2.95265 - the article this story is about — the whole article — the passage: "Deepfake technology represents one of the most disruptive applications of AI, enabling the creation of highly realistic synthetic audio-visual content. These systems rely on deep learning models that can replicate human facial expressions, voice patterns, and behaviors with high accuracy (Mirsky & Lee, 2021)."
  5. Kumari, A. (2026). AI and Cybersecurity: Combating Deepfakes and Misinformation in Nepal. Nepal Journal of Multidisciplinary Research. https://doi.org/10.3126/njmr.v9i2.95265 - the article this story is about — the whole article — the passage: "The advancement of generative adversarial networks (GANs) has further accelerated the development of deepfake technologies. These models operate through adversarial training, where two neural networks compete to improve the realism of generated content (Tolosana et al., 2021)."
  6. Kumari, A. (2026). AI and Cybersecurity: Combating Deepfakes and Misinformation in Nepal. Nepal Journal of Multidisciplinary Research. https://doi.org/10.3126/njmr.v9i2.95265 - the article this story is about — the whole article — the passage: "Additionally, deepfakes can influence political discourse by fabricating speeches and events involving public figures (Bennett & Livingston, 2021)."
  7. Kumari, A. (2026). AI and Cybersecurity: Combating Deepfakes and Misinformation in Nepal. Nepal Journal of Multidisciplinary Research. https://doi.org/10.3126/njmr.v9i2.95265 - the article this story is about — the whole article — the passage: "Algorithm-driven recommendation systems prioritize engagement-based metrics, often amplifying emotionally charged or sensational content regardless of its factual accuracy. This structural design enables misleading information to spread rapidly across interconnected user networks, often reaching large audiences before verification mechanisms can intervene (Cinelli et al., 2021; Zannettou et al., 2022)."
  8. Kumari, A. (2026). AI and Cybersecurity: Combating Deepfakes and Misinformation in Nepal. Nepal Journal of Multidisciplinary Research. https://doi.org/10.3126/njmr.v9i2.95265 - the article this story is about — the whole article — the passage: "Specifically, SMPM has a notable negative effect (β = -0.126, p = 0.041), implying that existing actions taken by social media platforms may be ineffective or even harmful in reducing misinformation within Nepal's context."
  9. Kumari, A. (2026). AI and Cybersecurity: Combating Deepfakes and Misinformation in Nepal. Nepal Journal of Multidisciplinary Research. https://doi.org/10.3126/njmr.v9i2.95265 - the article this story is about — the whole article — the passage: "The negative coefficient may not necessarily indicate that platform measures are objectively ineffective. Instead, it may reflect respondents' perceptions regarding the adequacy, transparency, consistency, or trustworthiness of current platform moderation practices."
  10. Kumari, A. (2026). AI and Cybersecurity: Combating Deepfakes and Misinformation in Nepal. Nepal Journal of Multidisciplinary Research. https://doi.org/10.3126/njmr.v9i2.95265 - the article this story is about — the whole article — the passage: "A further limitation of this study is that the quantitative design does not allow for an in-depth exploration of respondent's perceptions regarding social media platform measures."
  11. Kumari, A. (2026). AI and Cybersecurity: Combating Deepfakes and Misinformation in Nepal. Nepal Journal of Multidisciplinary Research. https://doi.org/10.3126/njmr.v9i2.95265 - the article this story is about — the whole article — the passage: "Digital literacy programs that focus on fact-checking, source verification, and critical thinking skills have shown positive effects in reducing susceptibility to manipulated content. Studies suggest that users with stronger media literacy competencies are more capable of identifying deceptive narratives and resisting misinformation exposure (Guess et al., 2021; Jones-Jang et al., 2021)."
  12. Hove MF, Hjorth F, Golovchenko Y, Adler-Nissen R, Bechmann A, van Dalen A, et al. (2026). Seeing Is Believing? Most People Struggle to Detect AI-Generated Images. Social Media + Society. 10.1177/20563051261483376 — only the abstract - the full text could not be fetched — the passage: "In a preregistered survey experiment, we find that, on average, people are not substantially better than chance at discerning AI-generated images from authentic images, regardless of the context of the image."
  13. Hove MF, Hjorth F, Golovchenko Y, Adler-Nissen R, Bechmann A, van Dalen A, et al. (2026). Seeing Is Believing? Most People Struggle to Detect AI-Generated Images. Social Media + Society. 10.1177/20563051261483376 — only the abstract - the full text could not be fetched — the passage: "While younger people, highly analytical-thinking individuals, and experienced users of generative AI have better discernment abilities, no demographic group other than young individuals (<25) performs substantially better than chance."
  14. Hussain M, Saeed F, Aldera S. (2026). Bioinspired Adaptive-Depth Neural Growth for Deepfake Video Forensics: An Entropy-Guided State-Space Framework. Biomimetics. 10.3390/biomimetics11090653 — only the abstract - the full text could not be fetched — the passage: "human evaluators accurately recognize high-quality forgeries merely 25% of the time, highlighting the pressing necessity for automated, widely applicable detection mechanisms"

Kumari, A. (2026). AI and Cybersecurity: Combating Deepfakes and Misinformation in Nepal. Nepal Journal of Multidisciplinary Research. https://doi.org/10.3126/njmr.v9i2.95265

Who paid: The article does not say who funded the study, whether any funder had a say, or who lent equipment or software.

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

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