The age for using AI in class depends on three different rules: those of the companies, the law and the education authorities
Each service’s terms, data protection law and some education authorities set different ages for using artificial intelligence, and they are often confused. This report explains what each says as of 6 November 2026, why research asks us to swap the question of age for the question of purpose, and how a primary or secondary teacher can decide which tool to use, with which account, and who operates it.
9 minute read
Three rules that get confused
When a teacher wonders whether their students can use an artificial intelligence assistant, they usually look for a number. But at least three rules of different origin coexist. The first is the minimum age each company sets in its terms of use, a contract between the company and whoever opens the account. The second is the age from which a minor can consent by themselves to the processing of their data, which Spanish law sets at 14. The third is the recommendations of education authorities, which guide what is considered prudent in a classroom.
The three answer different questions and do not coincide. That a service admits accounts from the age of 13 does not mean it is suitable for a Year 8 class, or that a school can ask its students to sign up for it. And that a student has turned 14 does not settle whether the tool has an educational purpose. Besides the students’ age, what counts is who opens the account, who operates the tool and what it is used for.
What companies set
As of 6 November 2026, OpenAI’s terms of use for Europe require users to be at least 13 or the minimum age their country requires to consent, and ask for the permission of a parent or legal guardian for those under 18. Its help centre (in Spanish) adds a note aimed at the classroom: if ChatGPT is used for educational purposes with children under 13, the direct interaction with the tool must be carried out by an adult.
Google distinguishes two worlds. With a personal account, its help pages say that in Spain you need to be at least 14 to manage your own Google Account; the Gemini apps ask for 13 or the age set by each country, but they are not available for supervised accounts with Family Link in the European Economic Area, and to use them with a paid AI plan in the European Economic Area you must be over 18. With a Google Workspace for Education account, by contrast, the guide for schools (in Spanish), updated on 28 October 2026, presents Gemini and Gemini Notebook as available to all ages, with access managed by the administrator, more heavily filtered content for under-18s and data that is not used to train models.
On those accounts, according to the age-based access control page (in Spanish), in primary and secondary schools anyone not identified by the administrator as over 18 gets a restricted experience: for example, they cannot use Gemini in Classroom to create or edit content. In addition, according to Engadget, Google announced that Gemini and search will be switched off by default in primary and secondary school accounts until the school turns them on, with a rollout “in the coming months”; the newspaper covered the debate that announcement sparked about who decides in a school.
Microsoft sets a minimum age of 13 in the Copilot terms for individuals, which apply from 18 August 2026, and which may be higher depending on each country’s law, and it warns that it may switch off or limit features for under-18s. For schools, its education blog announced in August 2025 Microsoft 365 Copilot Chat for students aged 13 and over, managed by administrators and without using the data to train foundation models. Anthropic is the most restrictive: the Claude consumer terms, in force since 8 October 2025, require users to be at least 18.
The law: fourteen, and the educational function
Article 7 of the Data Protection Act (in Spanish) establishes that the processing of a minor’s data can be based on their consent only when they are over fourteen; below that, it is lawful only with the consent of the holder of parental authority or guardianship, as the Spanish Data Protection Agency (in Spanish) reminds us. It is the age that counts when a student opens a commercial service on their own.
At school, however, consent is not the usual basis. The AEPD explains that in compulsory and regulated education neither students’ nor families’ consent is needed (in Spanish) for processing linked to the educational function, because it answers a public-interest mission set out in the Education Act. The platforms are chosen by the school or the administration, with the advice of its data protection officer, and teachers must use the means these put at their disposal. The difference is clear: using a tool contracted by the education authority is not the same as asking a 12-year-old to register for a commercial service.
That age of 14 could have changed. The Bill to protect minors in digital environments (in Spanish), introduced in March 2025, included raising it to 16. According to its page on the Congress website (in Spanish), consulted on 6 November 2026, the initiative is listed as lapsed since 6 October 2026, the date the Cortes were dissolved. The age in force therefore remains 14.
The European framework adds two pieces. The AI Act (in Spanish) prohibits systems that exploit vulnerabilities arising from age and classes as high-risk those used in education to decide access or assess outcomes. According to the European Commission, the AI literacy obligation has applied since 2 February 2025, and the high-risk obligations of Annex III have been postponed to 2 December 2027.
What education authorities say
Each autonomous community decides how artificial intelligence is used in its schools, and as of 7 November 2026 there is no common criterion. Some have published a document of their own, others handle it through the settings of their platforms, and several have not yet said anything.
Catalonia has the most detailed criterion. Its Orientacions per a l’ús de la intel·ligència artificial als centres educatius, published by the Department of Education and Vocational Training in September 2026 and later removed from its website (the newspaper’s review in its Library), state that students under 14 should not use general-purpose AI autonomously: in primary school and the first cycle of ESO, if they work with it, the teacher operates it, although tools designed for the educational setting that meet certain requirements are allowed. Between 14 and 18, families’ consent is not, as a general rule, needed, because the legal basis is the public-interest mission, but they recommend collecting their authorisation as a safeguard, and it may be necessary because some tools require it in their terms; with that authorisation, they place students’ direct interaction from the second term of Year 9 (3.º de ESO).
The guidance from the Basque Government’s Department of Education, from 2025, does not allow chatbots for under-13s or in the first two years of ESO; from the third year, it only allows them when necessary and under strict supervision. Between 14 and 18 it recommends using them without registering, because registering would require families’ authorisation.
In the Region of Murcia, an order of September 2025 does not allow, until the third year of ESO, activities involving direct access to AI tools that collect personal data or require autonomous interaction: before then, only guided use of age-appropriate apps, and from then on, always under teacher supervision. Families must be informed, and the order does not require their authorisation. In the Community of Madrid, the Política de uso aceptable de la IA, a proposal signed by the Regional Ministry for Digitalisation on 25 September 2026 that will come into force once approved, reserves the use of AI for teachers and school leadership teams; when the legal basis is consent, the family signs for under-14s. The same ministry’s Google Workspace guidelines state that students only use Gemini guided by the teacher and without their own credentials.
Other authorities handle it through their platforms and instructions. Andalusia has Gemini and NotebookLM switched off for students in its schools’ Google accounts, and the Canary Islands open NotebookLM to students and reserve Gemini for teachers. In the Valencian Community, the guide for planning teacher training on AI recalls that only apps authorised in the AppsEdu catalogue may be used with students and that they should not be asked to create accounts on unauthorised external services. Extremadura’s Instrucción 6/2026 on digital devices states that activities with school equipment do not require specific family authorisation and that students should declare when they use AI if the activity requires it. Asturias forbids entering students’ data into AI tools not authorised by its Regional Ministry (start-of-year circular).
Aragon, the Balearic Islands, Cantabria, Castile and León, Castile-La Mancha, Navarre and La Rioja do not yet have a document of their own on AI in schools; Castile and León announced a guide of recommendations in October that it has not yet published, and La Rioja publishes the INTEF guide on its website. Galicia will regulate it in its Digital Education Act, whose text is going through its Parliament as a bill. In Ceuta and Melilla, whose schools are run by the Ministry, the instructions for the school year do not deal with AI.
At national level, the INTEF’s Guía sobre el uso de la inteligencia artificial en el ámbito educativo, version 2.0, published on 8 September 2026, does not set an age. It points out that under 14 any registration requires families’ consent, proposes activities that require the teacher’s mediation and asks that only tools validated by the authorities or authorised by the school’s management be used with students.
What each education authority says, as of 7 November 2026
Authority
Age
Family authorisation
Document
Andalusia
No age set. Gemini and NotebookLM switched off for students.
Where the authority has published nothing or does not deal with authorisation, the general rule for all of Spain applies. With tools provided by the school or the authorities, families’ consent is not needed, because the processing is based on the educational function, but they must be told which tool is used, with which account and for what. If the activity requires students to register or relies on consent, under 14 the family authorises it, and from 14 students can do so themselves. Family authorisation is also needed when the tool’s terms require it: OpenAI, for example, requires it up to the age of 18. The INTEF guide serves as a reference: it proposes activities mediated by teachers and only tools validated by the authorities or authorised by the school’s management.
Beyond the number
UNESCO put a figure on it in 2023: its guidance on generative AI in education proposed a minimum age of 13 for using these tools in the classroom. Three years later, a research report published by UNESCO itself, written by Sonia Livingstone, Kim R. Sylwander, Sandra El Gemayel and Ayça Atabey, argues that there is no right age, and the newspaper covered it in its Library. In her summary on the Media@LSE blog, Livingstone explains that developmental science now sees variability and context as the norm, and that without attention to child development and clear educational purposes, ministries risk deciding without foundation.
The report also proposes distinguishing between educating students about AI, a literacy that should start early, and using AI to educate, where the evidence of benefits is scarce. That is why Livingstone considers the precautionary principle “a reasonable starting point”. She also cites Common Sense Media: 73% of American teenagers say they have never talked to a teacher about what AI is and how it works. The summary does not give the sample size.
The World Economic Forum, in a white paper from September 2026, offers the view opposed to broad bans. It argues that age verification is useful for applying differentiated safeguards, but does not determine maturity or guarantee moderation or redress for harm, and that blanket restrictions can block benefits and push children towards less controlled services. The two positions share the substance: age alone does not make a tool safe or useful.
What a teacher can do
With students under 14, the most prudent option, and the one consistent with what the authorities that have spoken out ask for, is mediated use: the teacher operates the tool with their school account, projects and comments, and students analyse, check and discuss the answers without entering personal data. This way AI literacy can be taught from primary school, for example by checking a chatbot’s errors or comparing its answer with a book, without any child opening an account.
When students are going to use the tool directly, it is best to do so with the educational account provided by the administration and not a personal one, after checking that the school’s administrator has enabled it and with what limitations. A personal account opened by a minor falls outside the school’s umbrella and depends on consent and the company’s terms. If in doubt about a particular tool, the education authority’s data protection officer can advise.
Families must be informed about which tool is used, with which account and for what, even if the law does not always require their consent. Before each activity, the questions raised by the research help: what educational purpose does it serve, what evidence supports it, and is it about learning about AI or with it. And it is worth reviewing the services’ terms from time to time, because they change often.
The age for using AI in class depends on three different rules: those of the companies, the law and the education authorities
Each service’s terms, data protection law and some education authorities set different ages for using artificial intelligence, and they are often confused. This report explains what each says as of 6 November 2026, why research asks us to swap the question of age for the question of purpose, and how a primary or secondary teacher can decide which tool to use, with which account, and who operates it.
Three rules that get confused
When a teacher wonders whether their students can use an artificial intelligence assistant, they usually look for a number. But at least three rules of different origin coexist. The first is the minimum age each company sets in its terms of use, a contract between the company and whoever opens the account. The second is the age from which a minor can consent by themselves to the processing of their data, which Spanish law sets at 14. The third is the recommendations of education authorities, which guide what is considered prudent in a classroom.
The three answer different questions and do not coincide. That a service admits accounts from the age of 13 does not mean it is suitable for a Year 8 class, or that a school can ask its students to sign up for it. And that a student has turned 14 does not settle whether the tool has an educational purpose. Besides the students’ age, what counts is who opens the account, who operates the tool and what it is used for.
What companies set
As of 6 November 2026, OpenAI’s terms of use for Europe require users to be at least 13 or the minimum age their country requires to consent, and ask for the permission of a parent or legal guardian for those under 18. Its help centre (in Spanish) adds a note aimed at the classroom: if ChatGPT is used for educational purposes with children under 13, the direct interaction with the tool must be carried out by an adult.
Google distinguishes two worlds. With a personal account, its help pages say that in Spain you need to be at least 14 to manage your own Google Account; the Gemini apps ask for 13 or the age set by each country, but they are not available for supervised accounts with Family Link in the European Economic Area, and to use them with a paid AI plan in the European Economic Area you must be over 18. With a Google Workspace for Education account, by contrast, the guide for schools (in Spanish), updated on 28 October 2026, presents Gemini and Gemini Notebook as available to all ages, with access managed by the administrator, more heavily filtered content for under-18s and data that is not used to train models.
On those accounts, according to the age-based access control page (in Spanish), in primary and secondary schools anyone not identified by the administrator as over 18 gets a restricted experience: for example, they cannot use Gemini in Classroom to create or edit content. In addition, according to Engadget, Google announced that Gemini and search will be switched off by default in primary and secondary school accounts until the school turns them on, with a rollout “in the coming months”; the newspaper covered the debate that announcement sparked about who decides in a school.
Microsoft sets a minimum age of 13 in the Copilot terms for individuals, which apply from 18 August 2026, and which may be higher depending on each country’s law, and it warns that it may switch off or limit features for under-18s. For schools, its education blog announced in August 2025 Microsoft 365 Copilot Chat for students aged 13 and over, managed by administrators and without using the data to train foundation models. Anthropic is the most restrictive: the Claude consumer terms, in force since 8 October 2025, require users to be at least 18.
The law: fourteen, and the educational function
Article 7 of the Data Protection Act (in Spanish) establishes that the processing of a minor’s data can be based on their consent only when they are over fourteen; below that, it is lawful only with the consent of the holder of parental authority or guardianship, as the Spanish Data Protection Agency (in Spanish) reminds us. It is the age that counts when a student opens a commercial service on their own.
At school, however, consent is not the usual basis. The AEPD explains that in compulsory and regulated education neither students’ nor families’ consent is needed (in Spanish) for processing linked to the educational function, because it answers a public-interest mission set out in the Education Act. The platforms are chosen by the school or the administration, with the advice of its data protection officer, and teachers must use the means these put at their disposal. The difference is clear: using a tool contracted by the education authority is not the same as asking a 12-year-old to register for a commercial service.
That age of 14 could have changed. The Bill to protect minors in digital environments (in Spanish), introduced in March 2025, included raising it to 16. According to its page on the Congress website (in Spanish), consulted on 6 November 2026, the initiative is listed as lapsed since 6 October 2026, the date the Cortes were dissolved. The age in force therefore remains 14.
The European framework adds two pieces. The AI Act (in Spanish) prohibits systems that exploit vulnerabilities arising from age and classes as high-risk those used in education to decide access or assess outcomes. According to the European Commission, the AI literacy obligation has applied since 2 February 2025, and the high-risk obligations of Annex III have been postponed to 2 December 2027.
What education authorities say
Each autonomous community decides how artificial intelligence is used in its schools, and as of 7 November 2026 there is no common criterion. Some have published a document of their own, others handle it through the settings of their platforms, and several have not yet said anything.
Catalonia has the most detailed criterion. Its Orientacions per a l’ús de la intel·ligència artificial als centres educatius, published by the Department of Education and Vocational Training in September 2026 and later removed from its website (the newspaper’s review in its Library), state that students under 14 should not use general-purpose AI autonomously: in primary school and the first cycle of ESO, if they work with it, the teacher operates it, although tools designed for the educational setting that meet certain requirements are allowed. Between 14 and 18, families’ consent is not, as a general rule, needed, because the legal basis is the public-interest mission, but they recommend collecting their authorisation as a safeguard, and it may be necessary because some tools require it in their terms; with that authorisation, they place students’ direct interaction from the second term of Year 9 (3.º de ESO).
The guidance from the Basque Government’s Department of Education, from 2025, does not allow chatbots for under-13s or in the first two years of ESO; from the third year, it only allows them when necessary and under strict supervision. Between 14 and 18 it recommends using them without registering, because registering would require families’ authorisation.
In the Region of Murcia, an order of September 2025 does not allow, until the third year of ESO, activities involving direct access to AI tools that collect personal data or require autonomous interaction: before then, only guided use of age-appropriate apps, and from then on, always under teacher supervision. Families must be informed, and the order does not require their authorisation. In the Community of Madrid, the Política de uso aceptable de la IA, a proposal signed by the Regional Ministry for Digitalisation on 25 September 2026 that will come into force once approved, reserves the use of AI for teachers and school leadership teams; when the legal basis is consent, the family signs for under-14s. The same ministry’s Google Workspace guidelines state that students only use Gemini guided by the teacher and without their own credentials.
Other authorities handle it through their platforms and instructions. Andalusia has Gemini and NotebookLM switched off for students in its schools’ Google accounts, and the Canary Islands open NotebookLM to students and reserve Gemini for teachers. In the Valencian Community, the guide for planning teacher training on AI recalls that only apps authorised in the AppsEdu catalogue may be used with students and that they should not be asked to create accounts on unauthorised external services. Extremadura’s Instrucción 6/2026 on digital devices states that activities with school equipment do not require specific family authorisation and that students should declare when they use AI if the activity requires it. Asturias forbids entering students’ data into AI tools not authorised by its Regional Ministry (start-of-year circular).
Aragon, the Balearic Islands, Cantabria, Castile and León, Castile-La Mancha, Navarre and La Rioja do not yet have a document of their own on AI in schools; Castile and León announced a guide of recommendations in October that it has not yet published, and La Rioja publishes the INTEF guide on its website. Galicia will regulate it in its Digital Education Act, whose text is going through its Parliament as a bill. In Ceuta and Melilla, whose schools are run by the Ministry, the instructions for the school year do not deal with AI.
At national level, the INTEF’s Guía sobre el uso de la inteligencia artificial en el ámbito educativo, version 2.0, published on 8 September 2026, does not set an age. It points out that under 14 any registration requires families’ consent, proposes activities that require the teacher’s mediation and asks that only tools validated by the authorities or authorised by the school’s management be used with students.
What each education authority says, as of 7 November 2026
Authority
Age
Family authorisation
Document
Andalusia
No age set. Gemini and NotebookLM switched off for students.
Where the authority has published nothing or does not deal with authorisation, the general rule for all of Spain applies. With tools provided by the school or the authorities, families’ consent is not needed, because the processing is based on the educational function, but they must be told which tool is used, with which account and for what. If the activity requires students to register or relies on consent, under 14 the family authorises it, and from 14 students can do so themselves. Family authorisation is also needed when the tool’s terms require it: OpenAI, for example, requires it up to the age of 18. The INTEF guide serves as a reference: it proposes activities mediated by teachers and only tools validated by the authorities or authorised by the school’s management.
Beyond the number
UNESCO put a figure on it in 2023: its guidance on generative AI in education proposed a minimum age of 13 for using these tools in the classroom. Three years later, a research report published by UNESCO itself, written by Sonia Livingstone, Kim R. Sylwander, Sandra El Gemayel and Ayça Atabey, argues that there is no right age, and the newspaper covered it in its Library. In her summary on the Media@LSE blog, Livingstone explains that developmental science now sees variability and context as the norm, and that without attention to child development and clear educational purposes, ministries risk deciding without foundation.
The report also proposes distinguishing between educating students about AI, a literacy that should start early, and using AI to educate, where the evidence of benefits is scarce. That is why Livingstone considers the precautionary principle “a reasonable starting point”. She also cites Common Sense Media: 73% of American teenagers say they have never talked to a teacher about what AI is and how it works. The summary does not give the sample size.
The World Economic Forum, in a white paper from September 2026, offers the view opposed to broad bans. It argues that age verification is useful for applying differentiated safeguards, but does not determine maturity or guarantee moderation or redress for harm, and that blanket restrictions can block benefits and push children towards less controlled services. The two positions share the substance: age alone does not make a tool safe or useful.
What a teacher can do
With students under 14, the most prudent option, and the one consistent with what the authorities that have spoken out ask for, is mediated use: the teacher operates the tool with their school account, projects and comments, and students analyse, check and discuss the answers without entering personal data. This way AI literacy can be taught from primary school, for example by checking a chatbot’s errors or comparing its answer with a book, without any child opening an account.
When students are going to use the tool directly, it is best to do so with the educational account provided by the administration and not a personal one, after checking that the school’s administrator has enabled it and with what limitations. A personal account opened by a minor falls outside the school’s umbrella and depends on consent and the company’s terms. If in doubt about a particular tool, the education authority’s data protection officer can advise.
Families must be informed about which tool is used, with which account and for what, even if the law does not always require their consent. Before each activity, the questions raised by the research help: what educational purpose does it serve, what evidence supports it, and is it about learning about AI or with it. And it is worth reviewing the services’ terms from time to time, because they change often.
Key points
Three different ages coexist: the one in each service’s terms, the age of consent for data processing (14 in Spain) and the one some education authorities set.
As of 6 November 2026, OpenAI and Microsoft ask for at least 13, Anthropic 18, and Google opens Gemini to all ages only in educational accounts managed by the school.
In regulated education, data processing rests on the educational function, not on consent, as long as the tools provided by the school or the authority are used.
Each region decides: a few set ages or conditions, others handle it through their platforms and seven have no document of their own.
The bill that would have raised the age of consent to 16 lapsed when the Cortes were dissolved in October 2026.
UNESCO and the World Economic Forum agree that age alone is not enough: purpose, conditions and evidence matter.
