AI Chatbots Are Becoming Part of Childhood. Europe Is Starting to Ask Who Keeps Them Safe

Standfirst

AI is no longer a future technology for children. It is already part of homework, entertainment, curiosity and conversation. Irish research published this month found generative-AI use among children rising sharply, while Europe’s proposed KIDS Act explicitly targets risky digital services and AI systems. The next online-safety challenge is therefore bigger than social media: how do we make conversational AI safer without turning childhood into an identity-surveillance system?

Key takeaways

• CyberSafeKids’ 2026 research, reported by RTÉ, found generative-AI use among surveyed children had grown substantially: 55% of 8–12-year-olds and 76% of 12–15-year-olds were using AI.
• The European Commission’s proposed EU KIDS Act, published on 17 September 2026, is designed to protect minors from risky digital services and AI systems, not only traditional social media.
• Europe’s age-verification direction is increasingly privacy-preserving: the Commission says its solution can prove an age threshold without disclosing a person’s exact age, identity or other personal details.
• Ireland’s latest evidence also shows why education matters: 69% of children in Coimisiún na Meán’s national research preferred being taught how to stay safe online to a social-media ban.
• Age assurance can help establish age-appropriate experiences, but it cannot replace safe product design, digital literacy, effective reporting and enforcement.

AI has quietly become part of childhood

For years, the dominant child-online-safety debate centred on social media. That framing is already becoming too narrow.

CyberSafeKids’ tenth annual report, The Year Little Changed, surveyed more than 7,000 children during the 2025–26 academic year. RTÉ reported that 55% of surveyed 8–12-year-olds and 76% of 12–15-year-olds were using generative AI. Older children were more likely to use it for information and schoolwork, while younger children were more likely to experiment or ask entertaining questions.

Those numbers matter because conversational AI is different from a conventional feed. A chatbot can respond personally, immediately and repeatedly. It can explain homework, invent stories, answer embarrassing questions and simulate conversation. The same qualities that make it useful can also make errors, inappropriate responses or over-trust more consequential for younger users.

This does not mean AI is inherently harmful to children. It means that child safety cannot be bolted on after the product has already become part of daily life.

The EU is widening the definition of online safety

On 17 September 2026, the European Commission published its proposal for the EU Keeping Internet Digital Spaces Accountable and Trustworthy Act — the EU KIDS Act.

The proposal has attracted attention for its social-media age rules, including a gradual approach to account access and a proposed EU-wide age of 15 for autonomous account creation on specified social networking services and video-sharing platforms. But one of its most important features is broader: the Commission says the proposal seeks to protect minors from risky digital services and AI systems.

That signals a shift in regulation. Children’s digital lives do not stop at the border between a social network, a game, a search engine and an AI assistant. Policy is beginning to recognise the same reality.

Safety by design is more important than a warning screen

The Commission describes safety by design as a central pillar of its new approach. That principle matters particularly for AI.

For a child-facing or child-accessible AI system, safety by design can mean thinking about foreseeable risks before the conversation starts: age-appropriate defaults, how the system handles sexual or violent requests, whether it encourages dependency, how it responds to signs of distress, what data it retains, whether reporting tools are understandable, and how easily a young user can reach a trusted adult or appropriate help.

No single safeguard solves all of these problems. A date-of-birth box is not a safety architecture. Neither is an age-verification screen. Effective protection is layered.

Age assurance is becoming more private — if implemented well

One of the strongest objections to online age checks is understandable: people should not have to surrender a passport or expose their identity every time they use an ordinary digital service.

Europe is trying to separate proof of age from proof of identity. The European Commission says its age-verification blueprint allows a person to demonstrate that they meet a required age threshold without revealing their exact age, identity or other personal details. A functionally ready solution became available in April 2026, and the Commission has urged Member States to make age-verification tools available by the end of 2026.

That architecture could become important well beyond adult-content gates. If a service needs to know that someone is above or below an age threshold, the privacy-preserving goal should be to disclose the minimum fact required — not build a new database of identities and browsing behaviour.

Privacy and safety are not opposites. Poor implementation can undermine both; careful implementation can reinforce both.

Ireland’s children are asking for skills as well as rules

Another important piece of evidence arrived this month from Coimisiún na Meán.

Its Online Safety & Children’s Digital Lives report, based on national surveys of children aged 9–17 and their parents or caregivers, found that 69% of children surveyed thought teaching children how to stay safe online would be better than a social-media ban. Sixty-seven per cent thought a ban would be ineffective because children would find a way around it.

This should not be interpreted as evidence that children need fewer protections. The same research records harmful and upsetting online experiences. Instead, it suggests that regulation and education should not be treated as competing ideas.

A child can be protected by safer defaults and still need the skills to recognise manipulation, misinformation and unreliable AI output. A parent can have meaningful controls and still need understandable information about what a service does. A platform can verify an age threshold and still fail if the experience behind that gate is poorly designed.

The chatbot problem is also a trust problem

Generative AI produces fluent language. Fluency can feel like authority, particularly to a younger person who may not yet have the experience to distinguish confidence from accuracy.

That creates a digital-literacy challenge beyond simply spotting fake news. Young users need to understand that an AI response can sound certain and still be wrong; that a chatbot is not necessarily a confidential friend, doctor, teacher or counsellor; that personal information shared in a conversation may be processed according to the service’s policies; and that high-stakes advice should be checked with a trustworthy human or authoritative source.

These are learnable skills. Schools, parents, technology companies and public bodies all have a role in teaching them.

Ireland is already testing whether online-safety rules work in practice

The regulatory environment is moving from rule-writing toward enforcement.

On 8 September 2026, Coimisiún na Meán opened its first formal investigation under Ireland’s Online Safety Code, examining X’s implementation of age-assurance mechanisms and parental controls. The opening of an investigation is not a finding that X breached the Code, but it demonstrates that regulators are increasingly examining whether safeguards actually work rather than merely whether they exist on paper.

Earlier in 2026, the European Commission also opened formal proceedings concerning Snapchat under the Digital Services Act, including scrutiny of age assurance and child-protection risks. Again, proceedings are investigations, not findings of liability.

For AI providers, the direction should be difficult to miss: a safety feature that exists but is ineffective, inaccessible or poorly integrated may not satisfy regulators — or families.

What this means for Safegram

Safegram’s direction is based on a simple principle: safer participation should be designed into the service rather than added only after something goes wrong.

Live Safegram functionality includes end-to-end encrypted chat, user and business verification, social and marketplace functionality through Safegram Exchange, creator/business tools, and account-level safety controls.

Safegram’s teen and family safety architecture is designed around stronger separation between teen and adult discovery and messaging, age-aware access, and family safety notifications without giving family members access to the content of private conversations. Individual elements of this architecture may evolve as implementation, testing and European requirements develop.

Safegram’s broader AI capabilities, including assistant and business-facing AI features, should be assessed feature by feature. A roadmap concept, beta function or planned safeguard should not be described as universally live until it is confirmed in the production product. Safegram’s aim is to apply the same privacy-first and safety-by-design principles as AI functionality expands.

The bigger question is no longer whether children will use AI

They already do.

The meaningful question is what kind of digital environment surrounds that use.

Europe is moving toward a model combining age-appropriate access, privacy-preserving age assurance, safety by design and enforcement. Irish children, meanwhile, are telling policymakers that they also want the knowledge to navigate digital spaces safely.

Those ideas belong together.

A safer digital childhood will not come from one age gate, one ban or one AI filter. It will come from products designed with children’s developmental needs in mind, privacy-respecting ways to establish age where necessary, credible enforcement, parents who have usable tools, and young people who understand the systems they are using.

FAQ

Are children in Ireland already using AI chatbots?

Yes. CyberSafeKids’ 2026 research, reported by RTÉ, found generative-AI use among 55% of surveyed 8–12-year-olds and 76% of 12–15-year-olds. The figures describe the surveyed population and should not automatically be treated as estimates for every child in Ireland.

Does the proposed EU KIDS Act cover AI?

Yes. The European Commission says the proposal seeks to protect minors from risky digital services and AI systems as part of a coherent EU framework for child online safety.

Is the EU KIDS Act already law?

No. It is a European Commission proposal published on 17 September 2026 and must proceed through the EU legislative process. Its final wording may change.

Will age verification mean children have to upload passports to every platform?

Not necessarily. The Commission’s privacy-preserving approach is designed to prove that a user meets an age threshold without revealing their exact age, identity or other unnecessary personal details.

Can age assurance make an AI chatbot safe by itself?

No. Age assurance can support an age-appropriate experience, but safety also depends on product design, model behaviour, data practices, reporting systems, parental tools, education and enforcement.

Why is digital literacy important with AI?

AI systems can generate convincing but inaccurate information. Children need skills to verify claims, protect personal information, recognise inappropriate interactions and know when to seek help or advice from a trusted person.

What is Safegram doing differently?

Safegram is building around verification, privacy and safety-oriented participation. Its teen/family architecture is designed to limit inappropriate adult-to-teen discovery and contact while preserving private conversation content. Safegram distinguishes between functionality that is live and features or safeguards that remain in development.

CTA

AI is becoming part of everyday digital life faster than rules, schools and families can adapt. Safegram News will continue tracking what changes in Ireland and Europe — and what those changes mean in practice for young people, parents, businesses and technology platforms.

Follow Safegram News for evidence-led coverage of online safety, digital trust, privacy, verification and safer online communities.

Source references

1. European Commission, “Proposal for EU KIDS Act — EU Keeping Internet Digital Spaces Accountable and Trustworthy”, 17 September 2026; updated 21 September 2026.

2. European Commission, “EU KIDS Act to restrict social media platforms’ access to children in the EU”, 17 September 2026; updated 22 September 2026.

3. European Commission, “The EU approach to age verification”, updated 22 September 2026.

4. European Commission, “Commission urges Member States to rollout EU age verification app”, 29 April 2026.

5. Coimisiún na Meán, “Online Safety & Children’s Digital Lives — research report published”, 10 September 2026.

6. Coimisiún na Meán, “Investigation commenced into X under Online Safety Code”, 8 September 2026.

7. RTÉ News, “Age assurance measures not stopping underage social media access — report”, 2 September 2026, reporting CyberSafeKids’ The Year Little Changed.

8. European Commission, “Commission investigates Snapchat’s compliance with child protection rules under the Digital Services Act”, 26 March 2026.

Image package

Hero filename: Article21_AI_Childhood_Hero_4K.jpg

Alt text: Irish parent and teenage child discussing an AI chatbot on a smartphone at a bright Dublin kitchen table.

Caption: AI is already part of children’s everyday digital lives. The policy challenge is increasingly about making those experiences safer by design.

Supporting image 1 filename: Article21_AI_Digital_Literacy_Dublin_4K.jpg

Alt text: Diverse Irish secondary-school students working together with laptops and a smartphone in a modern Dublin learning environment.

Caption: AI literacy is becoming a core online-safety skill: young users need to understand that fluent answers are not always accurate answers.

Supporting image 2 filename: Article21_Privacy_Preserving_Age_Assurance_4K.jpg

Alt text: Parent and teenager using a phone with a subtle abstract age-check shield visual, without identity documents, in an authentic Irish home.

Caption: Europe’s emerging age-assurance model aims to prove an age threshold while disclosing as little personal information as possible.

Supporting image 3 filename: Article21_Safety_By_Design_AI_4K.jpg

Alt text: Teenagers using digital devices beside Dublin’s River Liffey in a premium editorial image about AI and online safety.

Caption: Safety by design asks a different question: not only what young people should avoid, but what technology providers should build differently from the start.