Education & Intelligence
Protect young people.Keep learning open.
Teach judgement. Design safeguards.
Make room for discovery.
[Proposal]A better learning loop.
Human understanding at every step
[Proposal]Understand
Ask how it works.
Question the answer.[Proposal]Practise
Work with guidance.
Explain a choice.[Proposal]Evaluate
Try without help.
Check what lasts.
[Position]Young people deserve a place in the future they will inherit
Young people should have opportunities to understand intelligent systems, question their answers and learn to work with them. Protection should make those opportunities safe and meaningful. A policy that closes access to learning needs to explain why less restrictive protections would not work.
Our position is to support age-appropriate learning, require evidence of educational value and hold providers accountable for the environments they create. Some services or features may be unsuitable for children. That does not settle what should happen in a supervised lesson, a carefully designed tutor or a classroom discussion about how models work.
The question for policymakers is specific: which activity creates which risk, for which age group, and what response protects the child while preserving worthwhile learning?
[Proposal]Keep the learner at the centre.
[Evidence]What the European documents currently say
In her State of the Union address on 16 September 2026, the Commission President announced a differentiated social-media approach: no social media below 13; no personal accounts below 15; only parent-created and supervised mini-accounts from 13 to under 15, with restricted features and time; and safety-by-design duties for ages 15–18. She said the Commission would propose the EU Kids Act on 17 September. This is an official announcement, not an enacted restriction. [1].
The full legislative text, service definitions and exceptions still need to be checked when published. The speech alone does not establish whether, or how, educational tools and AI chatbots will be covered. The table distinguishes that announcement from existing law, guidance and Parliament’s earlier resolution.
A companion designed for sustained emotional engagement and a learning tool are different use cases, although one product can combine both. The European Parliament's research service identifies additional concerns associated with companions. Calling a product educational does not exempt its actual behaviour from scrutiny. [2].
| Document & date | Status | Relevance & limits |
|---|---|---|
| EU Kids Act · State of the UnionEuropean Commission President 16 September 2026 | Official announcement | The announced social-media approach uses age bands below 13, 13–under 15 and 15–18. The legislative proposal is expected on 17 September. The speech does not settle the legal treatment of educational tools or AI chatbots. [1] |
| Protection of minors onlineEuropean Parliament 26 November 2025 | Adopted resolution | Calls for a default age of 16 without parental consent for social media, video-sharing and risky companions, and a social-media minimum of 13. A non-legislative position; it is not the forthcoming Commission proposal or an enacted blanket ban on educational tools. [3] |
| Digital Services Act · Article 28EU legislature General application: 17 February 2024 | Applicable law | Covered online platforms accessible to minors must take proportionate measures for their privacy, safety and security. Article 19 contains an exemption for micro and small providers, except designated very large platforms. Not every educational tool is an online platform under the DSA. [4] |
| DSA guidance on minorsEuropean Commission 14 July 2025 | Guidance | Risk-based recommendations for covered platforms, including safeguards around integrated chatbots. Voluntary guidance used in assessing compliance; following it does not automatically establish compliance. It is not a separate general age ban. [5] |
| AI Act · Articles 5 and 50EU legislature Relevant duties: 2 February 2025 / 2 August 2026 | Applicable law · scope matters | Article 5(1)(a)–(b) prohibits specified manipulative practices and exploitation of vulnerabilities, including age, subject to behavioural-distortion and significant-harm conditions; these bans apply since February 2025. Article 50(1), applying since August 2026, requires notice of direct interaction with an AI system unless obvious, subject to its exceptions. The July 2026 amendments include other transitional rules; these are not universal deadlines for all AI Act obligations. [6] |
| GDPR · Articles 5 and 8EU legislature / national rules Applies since 25 May 2018 | Applicable law | Data minimisation is already required. Where processing relies on consent for an information-society service offered directly to a child, Article 8 uses age 16; national law may lower it to 13. Below the applicable age, parental authorisation is required. This is a consent rule, not a general minimum age for learning with technology. [7] |
This is a selected policy overview, not an exhaustive statement of applicable law. Service scope, exceptions and national rules must be checked for any actual deployment. Announcements do not create an educational exemption.
[Evidence]Better answers and better learning are different outcomes
In a randomised study of high-school mathematics in Türkiye, an unrestricted GPT-based interface improved assisted practice performance, but students subsequently performed worse on an assessment without that assistance. A version designed to guide learning largely removed that disadvantage; it did not establish an improvement in unaided performance. This was a study in one school and subject, not proof that every chatbot harms learning. [8] · [9].
A separate study involving 194 Harvard undergraduates found higher immediate post-lesson scores with a specially designed tutor than with the comparison active-learning classes. Its university setting and two physics topics limit what we can infer about younger children, long-term retention or other tools. [10].
These findings support a practical distinction: the design of assistance matters, and success while receiving help is insufficient evidence of lasting understanding. They do not establish a universal age at which every system becomes appropriate.
Age limits of this evidence. The two learning studies cited here concern secondary-school and university students; neither tests children under 13. We do not infer learning benefits for that age group from these studies. Our position for younger children rests on protection and design principles, not on demonstrated learning outcomes in the evidence cited here. This is not a claim that no other studies exist.
UNICEF's guidance brings both sides into the same framework: children's development and opportunities, alongside privacy, safety and accountable design. It is guidance, not an experimental finding. [11].
[Evidence]Two outcomes. Measure both.
Task performance
What can someone complete with support?
Independent understanding
What can someone explain and apply on their own?
[Debate]The strongest argument for restrictions deserves an answer
Children should not bear the burden of testing products whose risks they cannot reliably recognise. Incorrect advice, manipulation, exposure to inappropriate material and pressure to remain engaged cannot be addressed simply by telling a young person to be careful. Parents and teachers also have limited time and information.
A strong argument for age thresholds is enforceability: a general access rule may be easier to administer than checking the activity, product behaviour and supervision in each interaction. A second is precaution: potentially lasting developmental harm may justify caution when benefits of early access remain uncertain. The Türkiye study is also a warning that educational use and supervision alone do not establish that unrestricted assistance supports learning.
These are arguments to assess, not proof that every age threshold will be effective or proportionate. Responsibility belongs with providers as well as supervising adults. Some features should be restricted, and access may need to be suspended when risks cannot be adequately controlled.
Our unresolved question is enforcement. IGIQANT does not yet have a validated, scalable model for activity-based access. Any proposed educational exception would need a legally permitted scope, an identifiable responsible operator, enforceable product settings and auditable supervision. It cannot depend on a user merely describing an activity as educational.
A defensible restriction should identify the harm, explain its scope and specify how effectiveness will be reviewed. An advocate for wider access must show that safeguards can work at the intended scale. We propose examining both access rules and educational alternatives against those requirements.
[Proposal]Our proposals for access that supports development
[Proposal]Start with the activity
Distinguish demonstrations, supervised practice, independent accounts and ongoing emotional companionship. Observable check: before a proposed deployment, document enabled features, permitted activities and the person or organisation responsible for each permission.
[Proposal]Match the environment to the learner
Respect applicable age rules and service terms. Observable check: in a test environment using synthetic accounts, attempt an out-of-scope action and verify that it is blocked or escalated to the named operator. This alone would not validate age assurance or safety at scale.
[Proposal]Make understanding the goal
For a proposed learning evaluation, pair assisted practice with a comparable unassisted task. Observable check: report assisted task accuracy separately from unassisted performance and a later retention check, using criteria fixed before evaluation. Any research involving children would require appropriate ethical and legal arrangements; no such study is announced here.
[Proposal]Make existing duties auditable
Data minimisation and other duties already arise under applicable law; we do not claim them as IGIQANT inventions. Our proposed operational check is to document retained data, retention limits, deletion procedures and an accountable contact, then verify deletion and escalation in a synthetic test. The legal scope must be assessed for the actual service.
[Proposal]Protect access fairly
Consider learners without paid tools, reliable devices or an adult available to supervise. Observable check: test whether the same learning objective can be reached through a low-bandwidth or offline alternative and document additional cost or adult support required. Availability alone does not establish equal outcomes.
[Proposal]Review outcomes and change course
Track unassisted understanding, serious safety or privacy incidents, and operator review time. Decision rule for a proposed evaluation: set comparison margins and a review-time ceiling in advance; a serious incident triggers suspension and review, and results outside either limit block expansion pending redesign. These are proposed evaluation rules, not tested thresholds or a claim of regulatory compliance.
[Evidence]International policy is more complex than “progress versus bans”
The United States' executive order of 23 April 2025 promotes early exposure to intelligent technologies and training for educators. It provides a concrete example of an education initiative; it is not evidence that every tool or form of access is suitable for every child. [12].
China also distinguishes services. Rules effective from 15 July 2026 restrict virtual intimate-relationship services for minors and require guardian consent for other covered anthropomorphic services for under-14s. Their scope expressly excludes educational and similar services without sustained emotional interaction. This is a specific legal distinction, not a claim that those services are free from all regulation. [13].
Australia provides an implemented comparison. Since 10 December 2025, age-restricted social-media platforms must take reasonable steps to prevent under-16s from creating or keeping accounts. The obligation falls on covered platforms; children and parents are not penalised under this framework. Its scope and exclusions matter: this is not a blanket prohibition on learning with technology. Implementation alone does not establish effectiveness. [14].
Our interpretation: Policy comparisons should examine actual measures and outcomes. A country's ambition to develop technology does not, by itself, resolve questions about children's access.
[Proposal]What we propose for IGI
Within IGIQANT's work, IGI (Infinite Growing Intelligence) is a proposed design approach exploring how intelligent systems could develop alongside human purpose, with governed memory and accountable limits on autonomy. “Infinite” expresses an open-ended ambition, not a measured or guaranteed outcome.
In education, that purpose must include the learner's growing understanding. Memory should have clear limits. A system's authority should remain proportionate to its role, and a person or organisation must remain responsible for deployment.
We propose judging collaboration by what people can understand and do, including when assistance is removed. The evidence presented here does not validate an IGIQANT educational product. Our contribution is a position and a set of questions to investigate.
Young people deserve protection, opportunities to learn and a growing ability to make their own judgements. We should design for all three.
[Evidence]Read the evidence. Check the context.
- Commission addressEuropean Commission · Official address
- EPRS briefingEPRS · Policy briefing
- Parliament resolutionEuropean Parliament · Resolution summary
- DSA · Articles 19, 28, 93EUR-Lex · EU legislation
- Commission DSA guidanceEuropean Commission · Guidance
- AI Act · consolidated July 2026EUR-Lex · EU legislation
- GDPR · Articles 5, 8, 99EUR-Lex · EU legislation
- School mathematics studyPNAS · School mathematics study
- Authors' full paperStudy authors · Full paper
- University tutoring studyScientific Reports · University study
- UNICEF guidanceUNICEF · Policy guidance
- US education orderUnited States · Executive order
- Chinese regulator: Articles 2 and 14Chinese regulator · Official rules
- Australian eSafety regulatorAustralia · Official regulator
Evidence and policy status reviewed . Policy announcements, adopted measures, research findings and IGIQANT proposals have different status. The forthcoming EU proposal may change the policy context. Links are provided for verification; cited institutions do not endorse IGIQANT.