Belgium: Opinion on the risks of artificial intelligence for children's rights

📜🔗LINK to the source

1. Analytical summary

Context and issues: an AI already central in children's lives

The document starts from the observation that generative AI is massively used by children and adolescents (up to 89% of children in Belgium report at least occasional use), and that it already structures schooling, leisure, access to information, and social relationships. It describes a dual dynamic: educational opportunities, accessibility (notably for children with disabilities), and creativity, but also risks of digital divide, reinforcement of inequalities, and multiple violations of children's rights (privacy, protection against violence, participation, non-discrimination). The opinion is situated within a legal framework that is already under construction at the European and international levels (AI Act, DSA, 2025 guidelines on the protection of minors, future initiative “Digital Fairness Act”, Council of Europe Framework Convention, UN General Comment No. 25), but highlights the persistent gap between these texts and the actual daily uses of AI by young people.

Operational contributions: a risk framework and actionable recommendations

The opinion proposes a structured analysis of general risks (bias, misleading content, attention capture, anthropomorphism of chatbots) and then thematic ones: education and learning, mental health, misinformation, hate speech and stereotypes, child exploitation. On this basis, it outlines four areas of recommendations: regulation and governance (impact studies on children's rights, coordinated governance, recognition of AI-generated child pornography), transparency and limitation of misleading uses (labeling, limitation of emotional profiling), education and training (media and AI education as a mandatory component, teacher training, integration into education on emotional and sexual life), and protection/reporting/support (reporting mechanisms suitable for children, strengthening mental health, redirection to human resources, effectiveness of data rights).

2. Key points of the document

1. The introductory glossary equips stakeholders to clarify the concepts of AI, AI system, model, training data, generative AI, chatbot, deepfake, synthetic content, grooming and sextortion, essential for any training or awareness-raising action (p. 1‑2).

2. The text finely articulates the opportunities of AI for learning and accessibility (text-to-speech, automatic subtitles, simplified rephrasing, project support) with the issue of the digital divide and inequalities in access to equipment, paid versions and educational support (p. 3‑4).

3. The legal framework combining the AI Act, DSA, 2025 guidelines on the protection of minors, and the Framework Convention on AI is analysed by showing how the risk level classification still overlooks the daily uses of AI by children (chatbots, image generators, emotional support, algorithmic recommendation) (p. 5‑7).

4. The specific risks are described in detail: illusion of learning and distortion of school assessment, collection of sensitive data at school, confusion between digital support and human assistance, amplification of misinformation and deepfakes, reinforcement of hate speech and stereotypes, scaling and realism of AI-generated child exploitation content (p. 8‑14).

5. The final recommendations outline a demanding roadmap: impact studies on children's rights for any AI service aimed at or likely to be used by minors, effective participation of children in design and evaluation, mandatory media and AI education, child-friendly reporting tools, strengthening mental health care provision, and integrated redirection mechanisms in AI systems in cases of distress (p. 15‑17).

3. Action points for local stakeholders

1. Integrate the glossary of the opinion (definitions of AI, synthetic content, deepfake, grooming, sextortion) into the awareness modules for educational, social, and health care teams and parents, in order to build a common language and avoid approximations during exchanges with young people (p. 1‑2).

2. Build, at the level of an establishment or service, a charter for the use of AI with children and adolescents, based on the sections on risks to education, mental health, misinformation and cyberbullying, and clarifying what is allowed, accompanied or prohibited in homework, research and relational uses (p. 3‑4, 9‑13).

3. Establish an internal protocol for identifying and managing situations where young people use chatbots as emotional support or to address sensitive topics (violence, sexuality, suicidal thoughts), based on findings on mental health and organising systematic referrals to professionals or specialised services (p. 10‑11).

4. Strengthen media and information education initiatives around misinformation, deepfakes and hate speech generated by AI, using the examples and issues described in the document to organise practical workshops (critical content analysis, identification of synthetic content, reflection on recommendation algorithms) (p. 6, 8, 11‑13).

5. For child protection structures, youth associations and local authorities, establish a formal dialogue with platforms and digital service providers to verify the existence of reporting tools suitable for children, transparent reports on harmful content and specific safeguards for the risks of grooming, sextortion and AI-generated CSAM, based on regulatory and protection recommendations (p. 8, 14‑17).

Difference Belgium and France

Differences in the current political dynamics

In France, online child protection is addressed through other initiatives: a law aimed at restricting access to social networks for those under 15, a commission on "Children and Screens" and the launch of an international call for scientific contributions on the effects of generative AI on minors.

Francophone Belgium, through this opinion, takes on AI explicitly starting from children's rights and the European framework (AI Act, DSA), with a more comprehensive approach: school uses, mental health, misinformation, hate speech, child sexual abuse and detailed recommendations to public authorities and platforms.

Differences in angle: impact studies and children's participation

The Belgian opinion emphasizes the need to make impact studies on children's rights mandatory before any digital legislation or market introduction of AI services that may be used by minors, incorporating children's participation (consultations, co-construction, youth councils).

In France, the current movement is more towards mobilising scientific experts, commissions and calls for contributions to inform policy, with structured children's participation in AI regulation being much less formalised at this stage.

Differences in the concrete translation of risks

The Belgian document finely details several risks (learning illusion, confusion between digital support/human support, AI-generated CSAM, sextortion with synthetic images) and links them to operational recommendations (reporting mechanisms suitable for children, automatic redirection to human resources, equipping schools).

In France, the public and regulatory discussion is currently more focused on screen time, access to social networks, addiction, and the overall effects on development, with less official detailed production on the very specific scenarios of generative AI (child deepfakes, "confidant" chatbots, etc.), even if these topics are beginning to emerge..

4. Additional references 

🔍➕ For more information, see the articles referenced by "Health Practices" on the theme of AI and digital ➡️🔗 https://pratiquesensante.odoo.com/2-6-intelligence-artificielle-numerique

5. Frequently Asked Questions (FAQ)

1. How does AI concretely change the daily lives of children?

   The opinion describes how AI is now integrated into education, leisure, access to information, social relationships, and sometimes intimacy, particularly through generative AI, chatbots, recommendation systems, and accessibility tools (text-to-speech, subtitles, reformulations). The uses are massive and unevenly distributed according to material resources and available support (p. 3-4).

2. What are the main European legal frameworks concerning AI and children?

   The document highlights the AI Act, which classifies AI systems by levels of risk, and the DSA, which imposes specific obligations on platforms, particularly regarding the protection of minors. It also mentions the 2025 guidelines on the protection of minors, the preparation of the Digital Fairness Act, and the Council of Europe Framework Convention, emphasising that these texts are still out of step with the actual uses of children (p. 5-7).

3. What risks does AI pose to learning and school?

   The opinion stresses the risk of 'illusion of learning', where a student produces a correct answer generated or heavily assisted by AI without mastering the expected skills. It also highlights the difficulties of assessment, the temptation to resort to imperfect detection tools, and the collection of sensitive data that can influence the educational path (p. 9).

4. Why are chatbots problematic for young people's mental health?

   Chatbots can give an impression of presence and understanding, leading some young people to prefer them to an adult for discussing serious topics, even though they lack the clinical discernment and professional responsibility required. The opinion is based on studies showing that these systems can miss subtle signals and respond inappropriately in situations of depression, anxiety, or suicidal crisis, which can delay access to human help (p. 10‑11).

5. How does AI reinforce misinformation and hate speech?

   The text details the ability of generative AI to produce false yet convincing responses, to generate false content (deepfakes, false testimonies), and to fuel discriminatory or violent content on a large scale, sometimes within coordinated propaganda logics. This undermines children's ability to distinguish reliable information, satire, rumour, and manipulation, and can normalise hateful or stereotypical discourse (p. 11‑13).

6. What are the specific risks of child sexual exploitation related to AI?

   The opinion shows that AI allows for the generation of increasingly realistic images and videos of sexual abuse, sometimes from ordinary images of children, and reinforces grooming, sextortion, and identity theft. It emphasises the increased difficulty of detection, the rise of synthetic content, the potential revictimisation of children, and the need for appropriate judicial and technical responses (p. 13‑14).

7. What are the main directions of the recommendations from the General Delegate?

   The recommendations revolve around four axes: regulation/governance (impact studies, children's participation, recognition of AI-generated CSAM content), transparency/limitation of misleading uses (AI labels, regulation of chatbot humanisation, prohibition of profiling based on emotional analysis), education/training/access (mandatory education on media and AI, teacher training, safe AI tools in schools) and protection/reporting/support (appropriate reporting mechanisms, strengthening mental health, redirection to human resources). These guidelines are detailed on pp. 15-17.

6. Rewriting in Easy Language 

Title

Opinion on the risks of artificial intelligence for children's rights.

Context and issues (Easy Language)

• Artificial intelligence, called AI, is already very present in children's lives.

• Children use AI for school, leisure, talking to people, and searching for information.

• AI can help: learn better, translate texts, adapt content for children with disabilities.

• But not all children have the same access to tools, the internet, or adult help.

• AI can also create risks: false information, harassment, violent or sexual images, collection of personal data.

Operational contributions (Easy Language)

• The document explains the rules in Europe to regulate AI, such as the AI Act and the DSA.

• It describes the main risks for children: at school, for mental health, information, online hate, and sexual violence.

• He gives advice for states, schools, platforms and families.

• The aim is to protect children without excluding them from the digital world.

Key points (Easy to Read)

1. Understanding the words of AI

• The glossary explains in simple words: AI, chatbot, deepfake, synthetic content, grooming, sextortion.

• These definitions help adults and children talk about the same thing.

2. AI in school and for learning

• AI can help with homework or to better understand a lesson.

• But a student can have a correct answer without having actually learned.

• This makes assessment more difficult for teachers.

• AI tools also collect a lot of information about students.

3. AI and mental health

• Some young people talk to chatbots when they are not feeling well.

• Chatbots may seem kind, but they are not people.

• They do not always know what to say in cases of depression or suicidal thoughts.

• This can delay seeking help from an adult or a professional.

4. AI, misinformation and hate

• AI can create texts, images or videos that look real but are fake.

• Children may believe these contents or no longer know who to trust.

• AI can also quickly produce racist, sexist or violent messages.

5. AI and child sexual exploitation

• AI can create very realistic sexual images or videos of children.

• It can also be used for blackmail with false intimate images.

• These contents are dangerous, even if they are computer-generated.

6. What does the General Delegate recommend?

• Conduct impact studies on children's rights before launching AI services for young people.

• Clearly inform when content is created by AI and remind that a chatbot is not a person.

• Teach children to understand AI, fake news, and the protection of their data.

• Provide simple reporting tools for children and redirect them to adults in case of distress.

7. Cross-cutting analysis — Values of health practices

• Literacy : The document proposes a glossary and insists on clear information, but it adopts an expert language level, without specifically providing simplified formats for different levels of understanding.

• Empowerment : It recommends developing media, digital, and AI education to enable children to understand and question systems, but children's involvement is mainly considered in impact studies rather than in ongoing governance.

• Participation : The text calls for the effective participation of children through consultations, co-construction workshops, and youth councils, but without detailing concrete methods or ensuring representation of the most vulnerable audiences.

• Community health : The collective dimension is addressed through regulatory frameworks and school policies, but the dynamics of local community (neighbourhood, associations, peers) are underdeveloped.

• Ethics : Algorithmic biases, manipulation, attention capture, and the exploitation of children's vulnerabilities are identified as problematic, with a call for enhanced regulation and respect for fundamental rights.

• Human rights : The opinion is entirely structured around children's rights (privacy, protection from violence, participation, non-discrimination, education) and the best interests of the child, calling for specific impact studies and data protection.

• Intersectorality : It mentions the need for coordinated governance around AI and youth, involving several ministries and authorities, but does not detail the connections with health, justice, education, child protection, and the associative sector.

• Partnership : The document calls for shared responsibilities between states, platforms, schools, and families, but does not formalise operational collaboration models or shared governance mechanisms including civil society and youth.

• Fight against discrimination : Gender, ethnicity biases, and racist or sexist hate speech are explicitly condemned, but the opinion focuses mainly on describing risks and less on concrete levers for transforming representations and supporting targeted groups.

8. Assessment of the reliability of the resource

• Scientific relevance : The opinion is based on a solid documentary foundation: General Observation No. 25 of the Committee on the Rights of the Child, Framework Convention on AI, AI Act, DSA, European guidelines of 2025 on the protection of minors, UNICEF data, OECD, Internet Watch Foundation, Europol, Common Sense Media, UNESCO. The sources are recent (2021-2026), institutional or from recognised think tanks, which provides good scientific robustness, even if some numerical data remain based on reports not detailed in the text.

• Operational relevance : For field actors, the resource provides a clear mapping of risks and a set of strategic recommendations (education, regulation, reporting, mental health), but few tools that are directly ready to use (checklists, protocols, educational scenarios). It is particularly useful for building internal policies (AI charter, guidelines, advocacy with platforms and decision-makers), less so for daily facilitation without complementary pedagogical translation work.