1. How are children using AI?

AI tools can support a range of activities, including:

• Explaining difficult school topics in simpler language.
• Helping with brainstorming, writing, translation and creative projects.
• Supporting some communication and accessibility needs.
• Generating music, images, stories and other creative material.
• Practising languages or exploring subjects through interactive questions.

These uses can be valuable when children understand the limitations of the technology and have suitable support. An AI answer is not automatically correct, a generated image is not automatically authentic, and a chatbot is not a substitute for a trusted adult or qualified professional.

Children need opportunities to learn how AI works, question its answers and recognise when a task requires human help.

2. The risks: more than inappropriate content

Online safety discussions often focus on harmful images and videos. AI introduces additional risks that can be harder to recognise.

• Misleading information: AI can produce confident but incorrect explanations, fabricated sources or unsafe advice.
• Deepfakes and impersonation: tools can generate or alter images, video and audio to make it appear that a real person said or did something they did not.
• Bullying and humiliation: fabricated or manipulated content can be used to embarrass, threaten or harass a child.
• Privacy: children may share their names, school details, photographs, location or personal worries without understanding how that information may be processed.
• Emotional dependency: a chatbot designed to imitate friendship or respond affectionately may encourage some children to form an unhealthy attachment.
• Unsafe advice: a chatbot may respond poorly to questions about mental health, relationships, self-harm or other sensitive subjects.
• Manipulative design: recommendations, notifications and extended conversations may encourage more time online than a child intended.

These risks are not inevitable in every AI product. They depend on the system's design, the safeguards in place, the child's circumstances and how the service is used. Responsible providers should assess foreseeable risks before making a product available to children.

3. AI chatbots and emotional wellbeing

Children may turn to chatbots because they are easy to access, respond quickly or appear to listen without judgement. A chatbot can be useful for practising a language or exploring a school topic. However, a system that imitates a close friend, romantic partner or counsellor can create a different set of risks.

The UK Government's July 2026 consultation response describes concerns about emotional dependency, prolonged conversations and children relying on chatbots for mental health support. It says the Government intends to require regular breaks for under-18s using in-scope chatbots and to work with regulators on harmful or inaccurate mental health advice. It also says further action, potentially including restrictions on particular services, may be considered if other safeguards are insufficient.

These announced measures should be distinguished from rules already in force: the details and implementation of new requirements matter.

Children should be encouraged to speak to a trusted adult when they are frightened, distressed or dealing with something serious. AI systems should not discourage them from seeking real-world help or present themselves as a replacement for professional support.

4. Deepfakes, bullying and digital consent

Generative AI makes it easier to create images, audio or video that appear realistic. A young person's photograph may be manipulated, a voice may be imitated or a fake image may be circulated to humiliate someone.

Such material can cause harm even when it is fabricated. The person targeted may experience fear, embarrassment, reputational damage or pressure from peers. Sharing, forwarding or reposting the content can increase the harm.

Children need to understand that:

• Someone's photograph or voice should not be copied or manipulated in ways that violate their consent or dignity.
• A realistic-looking image or video is not proof that an event happened.
• Creating or sharing abusive or intimate synthetic material may have serious consequences and can engage criminal law.
• They should not forward humiliating or intimate content, even if it is circulating widely.
• They can ask a trusted adult for help without being blamed for someone else's abusive behaviour.

Schools, youth services and families should discuss digital consent, image-sharing, impersonation and respectful online behaviour before a harmful incident occurs.

5. Children's privacy and personal information

Children may not fully understand what happens when they upload a photograph, enter a personal story or ask a chatbot for help with a sensitive problem. Depending on the service, information may be stored, reviewed, shared with providers or processed for additional purposes.

Organisations offering online services to children should consider the UK GDPR, the Data Protection Act 2018 and the ICO's Age Appropriate Design Code where applicable.

The ICO advises organisations to take particular care with children's profiling and automated decisions. For services likely to be accessed by children, profiling should generally be switched off by default unless there is a compelling reason otherwise, and appropriate safeguards should be built in.

Good practice includes:

• Collecting only the personal information that is needed.
• Explaining privacy settings in language children can understand.
• Avoiding unnecessary collection of photographs, voice recordings and precise location information.
• Making clear whether conversations may be stored or used to improve a service.
• Providing suitable ways to report problems and seek help.
• Considering whether the service is genuinely appropriate for the ages of children likely to use it.

Parents and professionals should check the terms and privacy settings of a service rather than assume that every AI tool is suitable for children.

6. What does UK online safety regulation require?

The Online Safety Act 2023 places duties on in-scope online services. Ofcom explains that services covered by the relevant provisions must assess whether children are likely to access them and, where required, complete children's risk assessments and implement measures to protect children from harmful content.

AI-generated content shared through a regulated user-to-user service can fall within the same online safety framework as other user-generated content. However, the Act does not automatically cover every standalone chatbot or every way in which AI can cause harm. Ofcom has explained that the scope depends on how a service functions.

In July 2026, the Government announced further planned protections in response to its national consultation, including measures concerning chatbot breaks, harmful mental health advice and age-appropriate protections. The policy response describes intended next steps; it should not be treated as evidence that every proposed requirement is already in force.

For parents, schools and service providers, the practical lesson is to check current requirements for the particular service and keep up with regulatory updates.

7. Safeguarding: what should parents, schools and youth services do?

Safeguarding should not depend on children being able to identify every risk by themselves. Adults and organisations need clear procedures for prevention, reporting and responding to harm.

Practical steps include:

• Set clear boundaries: agree which AI tools may be used, for what purposes and with what level of supervision.
• Teach critical thinking: encourage children to verify important information and question content that seems shocking, manipulative or too good to be true.
• Protect personal information: explain why private details, passwords, intimate images and identifiable information should not be shared carelessly.
• Make reporting easy: children should know which adult they can approach and what will happen when they raise a concern.
• Respond without blame: if a child has been targeted, prioritise their safety and emotional wellbeing rather than shaming them.
• Preserve appropriate evidence: record relevant details and URLs where safe to do so, but do not download, forward or circulate illegal intimate imagery of children.
• Escalate serious concerns: use the organisation's safeguarding procedure and contact the police or relevant child-protection service where appropriate.

Safeguarding policies should be updated to address AI-generated abuse, impersonation, chatbot risks, personal-data exposure and the use of AI in education and youth work.

8. Disability inclusion: protect children without excluding them

AI can offer benefits to some disabled and neurodivergent children, including alternative ways to communicate, access information, organise tasks or explore creative interests. For some children, a well-designed tool may help them participate more independently.

However, an overly broad restriction can also remove useful support. Safety decisions should consider the purpose of a tool, the child's age and needs, the risks involved and the availability of appropriate supervision and safeguards.

Organisations should involve disabled children and their families in decisions that affect them, use accessible explanations and provide reasonable adjustments. They should avoid assuming that all children with the same diagnosis, disability or communication style need identical support.

Inclusion and safeguarding should work together: the aim is to reduce harm while preserving access to safe learning, communication and creativity.

9. Who is funding AI safety—and how should success be measured?

Investment in children's online safety can come from government programmes, regulators, schools, charities, research organisations and technology companies. Funding may support research, age-appropriate product design, education, safety testing, reporting tools and specialist support for children affected by online abuse.

Funding announcements alone do not prove that children are safer. The public should be able to ask what was delivered, who benefited and whether the intervention reduced harm.

Questions worth asking include:

• How much funding was committed, and how much was actually spent?
• Were children, parents, teachers and safeguarding specialists involved in designing the intervention?
• Was the technology independently tested for foreseeable risks?
• Are reporting and support routes accessible to disabled children and those with additional communication needs?
• Are providers transparent about safety failures and how they are corrected?
• Are outcomes measured through evidence of reduced harm and improved support, rather than simply the number of users or features launched?
• Who is accountable when safeguards fail?

Publicly funded safety initiatives should publish clear objectives and proportionate evidence of outcomes, while protecting children's privacy.

10. A responsible AI checklist for services used by children

1. Design for children's needs. Consider age, developmental stage, accessibility and foreseeable misuse before launch.
2. Assess risks before deployment. Identify harmful content, unsafe interactions, privacy risks and routes through which children could be targeted.
3. Use age-appropriate safeguards. Ensure settings, explanations and restrictions match the service's risk profile and applicable legal duties.
4. Make AI identity clear. Avoid misleading children into believing a chatbot is a human friend or qualified professional.
5. Protect personal data. Limit collection, use appropriate security and provide understandable privacy information.
6. Prevent harmful engagement patterns. Assess whether design features encourage excessive use, emotional dependency or inappropriate conversations.
7. Provide accessible reporting. Make it possible for children and trusted adults to report harmful content and receive appropriate responses.
8. Monitor and improve. Investigate incidents, evaluate safeguards and respond to new evidence of harm.
9. Keep people accountable. Assign responsibility for safety decisions and provide routes for escalation and redress.

11. The AM-MediaSpace perspective: safeguarding must keep pace with technology

At AM-MediaSpace, responsible AI means using technology as an assistant, protecting people's rights and maintaining meaningful human oversight. These principles are especially important when children and other vulnerable people are involved.

AI should create opportunities for children to learn, communicate and express their creativity. It should not be used to exploit their trust, expose them to avoidable harm or make it harder for them to seek help.

Safeguarding needs to be built into the way tools are designed, funded, introduced and reviewed—not added only after something goes wrong.

Conclusion: responsibility belongs to the whole system

Children should be able to benefit from AI without being left to manage its risks alone. Parents and carers need practical information, schools and youth services need clear safeguarding procedures, and technology providers need to take responsibility for the systems they create and operate.

Government and regulators must also make sure that protections keep pace with changing technology, and that commitments are followed by implementation, enforcement and evidence of results.

The goal is not to reject every new technology. It is to make sure children's safety, privacy, dignity and development remain more important than engagement metrics or commercial growth.

Sources and further reading

• UK Government — Growing up in the online world: government response (July 2026)
• UK Government — Growing up in the online world: progress statement (June 2026)
• Ofcom — Protection of children duties under the Online Safety Act
• Ofcom — AI chatbots and online regulation: what you need to know
• Information Commissioner's Office — Children's Code guidance and resources
• Information Commissioner's Office — Profiling children and automated decisions
• Department for Education — Keeping children safe in education
• NSPCC — Online safety advice and support
• Online Safety Act 2023
• Data Protection Act 2018