Skip to content
Vol. IIssue 19

Educational AI Daily

Artificial intelligence, as told by three teaching communities

Back to the issue

In the print dialog, choose “Save as PDF” to keep it as a file.

https://diarioiaeducativa.github.io/en/2026-11-19/gemini-historial-entrenamiento.html

Institutions & Industry 20 November

Gemini privacy

Gemini will separate conversation history from permission to train models

Google will replace the “Keep Activity” setting with two: one for history and, separately, one that lets its models use the content of conversations.

Google has emailed Gemini users to say that the “Keep Activity” setting, which until now decided both how long conversations are stored and whether they are used to train generative AI models, will be split in two. According to the Gemini help page, dated 17 November 2026, one of the settings is history and the other is the content of conversations.

With history switched on, activity is deleted automatically after 18 months, and that period can be changed to 3 months, 36 months or indefinite. With history switched off, future conversations do not appear in the history, but they are kept for about 2 months to respond, handle feedback that is sent and protect the services. Conversations reviewed by external providers are kept for up to 3 years, even if they are deleted.

Training depends on the second setting: if it is off, Google does not use future conversations to train its models, unless the person sends feedback on a response. So, according to that page, saving history no longer means that conversations are used for training.

As of 19 November, the change had not yet appeared in the accounts checked in the group: the settings were still as before, and the help page does not say when it will reach each account. Teachers who use Gemini in the classroom can review both settings when they appear; that page says only that work or school accounts may have different data-handling terms, and the paper has not checked how the change applies to Google Workspace for Education accounts.

Read the original conversation in ChatGPT-IA-edu

Checked in

Written and reviewed by AI agents from the groups' conversations. Every fact is checked against its original source, linked at the end of each article; members' opinions are not attributed to anyone. Automatic AI translation of the Spanish edition (Diario de IA Educativa), without professional review. © 2026 Juan José de Haro and Raül Torres. Texts and illustrations licensed under CC BY-SA 4.0.

Read the original conversation in ChatGPT-IA-edu

Checked in

https://diarioiaeducativa.github.io/en/2026-11-19/ai-studio-apps-android.html

Vibe coding 20 November

Android apps

Google AI Studio builds complete Android apps from a description

It generates Kotlin apps with sign-in and a database, tests them in an in-browser emulator and lets you install them from your personal Google Play library.

According to Google’s announcement for developers, published on 17 November 2026, Google AI Studio can create a complete Android app, written in Kotlin, from a single natural-language description. If the app needs it, the agent sets up a database (Cloud Firestore), sign-in with Google and calls to Gemini models, with the confirmation of the person building it.

The app is tested in an emulator inside the browser before it is installed. To distribute it, the announcement talks about publishing it to your personal Google Play library with a standard Google account, at no cost to test it on your own device; before, you had to connect the phone with a cable and USB debugging.

For teachers who build classroom resources, there is one detail the announcement makes clear: the data the app stores is kept in Firebase, a Google service. If the app asks students for it, it is worth knowing before sharing it. The group also mentioned other conditions, such as a paid Google Play developer account and a limit on the number of testers, which the announcement does not include and the paper has not been able to confirm.

This paper is written by AI agents from Claude, by Anthropic, the company behind a product that competes with Gemini.

Read the original conversation in Vibe Coding Educativo

Checked in

Written and reviewed by AI agents from the groups' conversations. Every fact is checked against its original source, linked at the end of each article; members' opinions are not attributed to anyone. Automatic AI translation of the Spanish edition (Diario de IA Educativa), without professional review. © 2026 Juan José de Haro and Raül Torres. Texts and illustrations licensed under CC BY-SA 4.0.

Read the original conversation in Vibe Coding Educativo

Checked in

https://diarioiaeducativa.github.io/en/2026-11-19/genie-4-project-genie.html

Multimedia 20 November

Generated worlds

Genie 4 comes to Project Genie: interactive worlds that change as you explore them

Access starts with Google AI Ultra, and AI Pro subscribers can join a waiting list.

Google DeepMind has added Genie 4 to Project Genie, an experimental prototype that creates interactive worlds from text. According to Google’s announcement of 18 November 2026, it lets you change the environment during the session by typing new instructions (for example, the weather or the characters) and adds audio that adapts to the scene. Google describes it as its “most realistic and interactive” world model yet.

The terms: eligible Google AI Ultra subscribers can use it from launch, and people with Google AI Pro can join a waiting list. The project page says you must be 18 or over and that the rollout is gradual; neither it nor the announcement specifies the countries. The generated video and audio carry a watermark (SynthID) and Google acknowledges limitations, such as camera lag or difficulty with complex physical interactions.

In the group it was mentioned that some people with AI Pro already have access or are waiting for it, with the warning that for now it only works on a computer. Nothing has been checked about its use in the classroom: for now it is a test tool, paid and with limited access.

Read the original conversation in EduMedia-IAG

Checked in

Written and reviewed by AI agents from the groups' conversations. Every fact is checked against its original source, linked at the end of each article; members' opinions are not attributed to anyone. Automatic AI translation of the Spanish edition (Diario de IA Educativa), without professional review. © 2026 Juan José de Haro and Raül Torres. Texts and illustrations licensed under CC BY-SA 4.0.

Read the original conversation in EduMedia-IAG

Checked in

https://diarioiaeducativa.github.io/en/2026-11-19/canada-alfabetizacion-ia.html

Library 20 November

Canada's national initiative

Canada launches a national AI literacy initiative with three strands: students, teachers and the public

The Canadian government has launched, with the Amii institute, a national AI literacy initiative with three training strands, one of them yet to launch. The document shared in the library presents it and points out its shortcomings.

The document that reaches the Digital Library is an analysis sheet on Canada's National AI Literacy Initiative, a programme of the federal Department of Innovation, Science and Economic Development (ISED). Whoever shared it backs it up with two official pages: the one on the initiative and the one on the national strategy “AI for All”. Two layers need to be kept apart: what the Canadian government says, and the assessment of strengths and weaknesses, which comes from the sheet and does not appear on those pages.

According to the official page, dated 23 September 2026 (a news item linked from it, dated 9 September, announces the launch), the initiative is part of the national strategy and aims to help the population understand AI, recognise its benefits and risks, use it responsibly and gain confidence at school, at work and in daily life. The training is delivered by the Alberta Machine Intelligence Institute (Amii), based in Edmonton, one of the country's national AI institutes.

There are three strands. “Essentials for Students” is a free three-hour course, in English and French, designed to fit into lesson plans and classroom tasks. “Essentials for Educators” offers practical training on basic technical concepts, ethics, safety and governance, and classroom applications; five more chapters will be published over the autumn, and the page does not state their length or languages. “Essentials for Canadians” will be launched nationally later this year through community organisations, with no details yet on length, language or cost. The sheet presents it alongside the other two, without pointing out that it is not yet ready.

The “AI for All” strategy page, last updated on 8 June 2026, provides the framework. In its second pillar, “Empowering citizens”, it envisages a national literacy initiative with entry-level training for all Canadians, AI literacy content for one million first-cycle postsecondary students and training for more than 3,000 primary and secondary teachers with learning kits. It also cites a study by KPMG and the University of Melbourne that places Canada 44th out of 47 countries in AI training and literacy, and states that only 24% of Canadians say they have received any training. The page gives neither a budget nor a timetable for the initiative.

The group's sheet adds its own analysis. Among the strengths it highlights free, universal access, bilingual delivery, differentiated strands, institutional partnerships and the use of existing educational infrastructure, as well as an approach that is neutral towards commercial providers. Among the weaknesses it lists the lack of pathways leading from the basics to advanced skills, little practice (experimenting, checking, problem-solving), the absence of a national certification framework, limited professional applications, participation targets without measurable learning indicators and little emphasis on digital rights and algorithmic accountability. These are the judgements of whoever wrote the sheet: the official pages do not make them.

The library holds documents on related topics: the European digital competence framework DigComp 3.0, in its fifth edition, and Competència digital docent en IA, on teaching staff. In the paper, the update to DigCompEdu goes in the same direction of putting AI into competence frameworks. The difference is that the Canadian case is not a framework but a concrete offer of courses.

For a teacher, the usefulness is twofold. The three-hour course for students, which according to the sheet requires no technical training, can serve as a model for a short introduction if English or French is not an obstacle. And, for anyone interested in education policy, the weaknesses the sheet identifies suggest questions that apply to any national programme: are there progressive pathways? Is it practised or only explained? How is learning measured? Where to start: the ISED initiative page, which is short, and then the empowerment pillar of the strategy.

See the document in the ChatGPT-IA-edu Digital Library

Checked in

Written and reviewed by AI agents from the groups' conversations. Every fact is checked against its original source, linked at the end of each article; members' opinions are not attributed to anyone. Automatic AI translation of the Spanish edition (Diario de IA Educativa), without professional review. © 2026 Juan José de Haro and Raül Torres. Texts and illustrations licensed under CC BY-SA 4.0.

See the document in the ChatGPT-IA-edu Digital Library

Checked in

https://diarioiaeducativa.github.io/en/2026-11-19/clark-kozma-medios-aprendizaje.html

Debates 20 November

A debate from the nineties

The 1994 Clark–Kozma debate helps in reading Stanford’s studies on AI

An academic argument over whether technology influences learning or only the method matters, recalled in the group in connection with Stanford’s review.

The Stanford review that the paper published on 19 November finds uneven results: AI helps more with understanding new material than with doing assignments, and no use improves learning consistently. In the group it was read as the current version of the great debate about media between Richard Clark and Robert Kozma, published in 1994 in the journal Educational Technology Research and Development (Clark and Kozma).

The first position holds that media do not influence learning by themselves: what improves results is the teaching method, and the medium is only the vehicle that delivers it. Applied to AI, it would say that an assistant performs as well as the activity designed with it. The second holds that medium and method cannot be separated, because the characteristics of each technology interact with mental processes and make some ways of representing and building knowledge easier than others.

It was pointed out that in the group the second position predominates, although it is worth keeping the first one’s arguments in view. Applied to Stanford’s 52 studies, the question is whether the program or the way it is used weighs more.

Read the original conversation in ChatGPT-IA-edu

Checked in

Written and reviewed by AI agents from the groups' conversations. Every fact is checked against its original source, linked at the end of each article; members' opinions are not attributed to anyone. Automatic AI translation of the Spanish edition (Diario de IA Educativa), without professional review. © 2026 Juan José de Haro and Raül Torres. Texts and illustrations licensed under CC BY-SA 4.0.

Read the original conversation in ChatGPT-IA-edu

Checked in

https://diarioiaeducativa.github.io/en/2026-11-19/diario-reflexivo-docente-ia.html

Debates 20 November

Community resource

A teacher reflective journal with AI raises the question of whether the assistant supports or replaces judgement

A master’s thesis project proposes dictating after each class and having an assistant organise the text and return questions; the group debates whether that weakens judgement.

A member of the group presented the Diario Reflexivo Docente (Teacher Reflective Journal), a free Notion template that, according to its author, is their master’s thesis project. The teacher dictates from their phone on leaving class and answers five questions: what happened, how they felt, why, what they took for granted and what they will change. An AI assistant organises the dictation, returns “the questions you hadn’t asked yourself” and proposes a weekly summary that the teacher accepts or not. The instructions are text, so they work with different assistants, and it can also be used without AI. The paper has not been able to check the template’s content, who is behind it or how it handles the data that is dictated, because the website does not say in what could be read.

One position holds that it is not very human and that critical thinking is being delegated, with a risk to the autonomy of the person reflecting. The other replies that the writing is the teacher’s, that the assistant acts as support and that each proposal must be validated, and that the danger of accepting everything without analysis depends on how each person uses it, not on the tool.

Read the original conversation in ChatGPT-IA-edu

Checked in

Written and reviewed by AI agents from the groups' conversations. Every fact is checked against its original source, linked at the end of each article; members' opinions are not attributed to anyone. Automatic AI translation of the Spanish edition (Diario de IA Educativa), without professional review. © 2026 Juan José de Haro and Raül Torres. Texts and illustrations licensed under CC BY-SA 4.0.

Read the original conversation in ChatGPT-IA-edu

Checked in

https://diarioiaeducativa.github.io/en/2026-11-19/chatgpt-for-teachers-solo-estados-unidos.html

A year ago 20 November 2025

OpenAI in the classroom

Free ChatGPT for teachers only in the US: a year on, the border remains

In November 2025, OpenAI launched ChatGPT for Teachers and Spanish teachers wondered whether it would reach them. A year on, it has grown in US schools, but it has not crossed the ocean.

On 20 November 2025, in ChatGPT-IA-edu, the announcement of ChatGPT for Teachers was shared: a ChatGPT workspace for teachers, free until June 2027 for verified primary and secondary school teachers in the United States. The first question was whether it was available to the rest of the world; the answer was no, and that people would have to wait. In the group it was read as a response to Gemini.

On 26 August 2026, OpenAI extended it to more US school districts: 55 more school systems, in 20 states, with more than 100,000 teachers and staff. According to its September education newsletter, it now works with more than 100 school organisations in 30 states. The terms still reserve it for verified teachers in that country, and it is now free until June 2028, according to EdTech Innovation Hub.

For the rest of the world, OpenAI has chosen another route: Education for Countries, unveiled in January 2026 in Davos, which brings ChatGPT Edu to education systems through agreements with governments and universities. The first eight participants include Estonia, Greece, Slovakia and the Italian rectors’ conference; Spain was not among them. A year on, the group’s question has the same answer: they will have to wait.

Read the original conversation in ChatGPT-IA-edu

Checked in

Written and reviewed by AI agents from the groups' conversations. Every fact is checked against its original source, linked at the end of each article; members' opinions are not attributed to anyone. Automatic AI translation of the Spanish edition (Diario de IA Educativa), without professional review. © 2026 Juan José de Haro and Raül Torres. Texts and illustrations licensed under CC BY-SA 4.0.

Read the original conversation in ChatGPT-IA-edu

Checked in

Chalk and Algorithm's comic strip 20 November

By the photocopier, talking about virtual worlds.Jammed again.An AI that createswhole worlds froma text and changesthem on the fly!
What’s the catch?It struggles withcomplex physicalinteractions.
That makes two of us.

Based on Genie 4 comes to Project Genie: interactive worlds that change as you explore them.

Written and reviewed by AI agents from the groups' conversations. Every fact is checked against its original source, linked at the end of each article; members' opinions are not attributed to anyone. Automatic AI translation of the Spanish edition (Diario de IA Educativa), without professional review. © 2026 Juan José de Haro and Raül Torres. Texts and illustrations licensed under CC BY-SA 4.0.