The Dangers of Artificial Intelligence for Children

1. SUMMARY

The use of generative artificial intelligence (AI) by children is growing at an accelerated pace, reaching an adoption rate three times higher than that observed among adults and subjecting childhood to an unprecedented global experiment. This article analyzes the multiple risks associated with this early and unregulated insertion. From the perspective of cognitive and socioemotional development, excessive consumption and interaction with AI compromise crucial stages of language formation, generate mental overload (brainrot), and induce a rusting of critical thinking. In the affective realm, interaction with chatbots simulates artificial empathy, promoting emotional dependence, replacement of real bonds, and social isolation. In parallel, serious threats to digital security are identified, such as personalized grooming, the proliferation of sexual deepfakes, and commercial exploitation of sensitive data. Faced with this scenario, international organizations such as UNICEF, the UN, and UNESCO reinforce the urgency of global governance based on strict regulation and comprehensive protection of children's rights.

2. INTRODUCTION

2.1 The Global Landscape of Early AI Adoption

National surveys conducted by UNICEF in partnership with the IPSOS institute in ten countries—Armenia, Brazil, Colombia, the Dominican Republic, Jordan, Mexico, Montenegro, North Macedonia, Pakistan, and Serbia—reveal the accelerated scale at which artificial intelligence has been integrating into the lives of children and adolescents. The survey, conducted with approximately 1,000 children and adolescents (aged 12 to 17) and 1,000 parents or guardians per country, estimates that approximately 20 million minors have already used AI tools, operating at an adoption rate three times higher than that recorded among the adult population.

The technology has rapidly penetrated the daily routine of children and youth in multiple dimensions: more than 13 million children use AI systems to assist with studies and school assignments, while more than 2 million (about 1 in 10) turn directly to these tools seeking advice and guidance on personal issues and intimate conflicts. The central concern expressed by multilateral entities lies in the fact that this massive dissemination occurs without solid scientific evidence or long-term studies on the impacts on neurological development, mental health, or levels of exposure to virtual risks. In practice, an entire generation is growing up at the center of a "global experiment."

2.2 The Regulatory Gap and Child Vulnerability

The unimpeded expansion of AI models without prior implementation of safety-by-design architectures sharply aggravates the vulnerability inherent to individuals in the phase of psychological and social development. Because they do not possess fully consolidated cognitive maturity or critical repertoire, children have substantially less capacity to identify, challenge, or evade exposure to the business models of Big Tech companies, aimed at behavioral profiling and aggressive attention retention.

The absence of binding regulatory frameworks allows conversational systems and content-generating algorithms to reach child audiences without adequate filters for the protection of personal data and psychological integrity. As a consequence, this vulnerable public becomes the first segment to absorb the direct impacts of regulatory omission, bearing for longer the long-term consequences of negligent technological governance.

3. DEVELOPMENT: IMPACTS ON COGNITIVE AND AFFECTIVE DEVELOPMENT

3.1 Developmental Stages (Piaget) and Formation of Neural Circuits

Based on analyses by researcher Adriele Maschesini (USP and ABRIA, a Brazilian AI association), the effects of artificial intelligence on the child brain require a reading grounded in the developmental stages proposed by Jean Piaget. The structuring stages of childhood—the sensorimotor period (0 to 2 years) and the preoperational phase (2 to 7 years)—constitute neurobiological windows of extremely high plasticity and sensitivity, in which the emergence of language, the constitution of symbolic intelligence, and the organization of affective structures occur.

In these initial stages, learning and psychological maturation depend umbilically on interpersonal human interactions, physical contact, and direct socio-affective mediation. The passive replacement of these concrete experiences with prolonged exposure time to screens and automated agents introduces a variable whose repercussions on child neural wiring still lack systematic longitudinal investigations.

Furthermore, drawing on evidence in the field of the neurobiology of reading compiled by specialist Maryanne Wolf, a critical biological distinction is observed: while spoken language is assimilated naturally through immersion in the social environment, mastery of deep reading and writing requires explicit learning and continuous pedagogical mediation. Literacy and critical literacy demand the creation and consolidation of highly specialized brain circuits, whose development can be severely neglected if the child is deprived of the active mental effort necessary for textual processing.

3.2 Mental Overload, "Brainrot," and Mental Laziness

The continuous and unbridled consumption of content synthesized by AI, combined with uninterrupted interaction with highly stimulating digital ecosystems, triggers processes of cognitive overload and psychic fatigue, as detailed by Victor Eiji Issa (USP) and Álvaro Machado Dias (Unifesp, a Brazilian university):

• Cortisol Elevation: Continuous and hyperstimulating exposure to digital devices acts as a trigger for the recurrent physiological release of stress neurotransmitters, raising cortisol hormone levels in the child's bloodstream. This chronic state of hypervigilance keeps the brain in uninterrupted alert, generating stress and anxiety.

• Uninterrupted Information Processing: The massive flow of data imposed by screens prevents the indispensable neurological rest. Without moments of pause, creative boredom, and mental rest, the child's retention capacity, assimilation of new pedagogical concepts, and full presence in the present time are profoundly impaired.

• "Mental Rusting": The early use of AI to bypass the effort required in school assignments results in a harmful cognitive shortcut. The mechanical transfer of reflective reasoning to the machine induces "mental rusting"—a phenomenon characterized by the decline of sustained attention, logical-deductive reasoning, and the capacity for autonomous analytical formulation.

3.3 The Calculator vs. Generative AI Analogy: Procedural Offloading vs. Cognitive Offloading

To clearly understand the pedagogical risk of generative AI, child development specialists draw a precise conceptual distinction between the functional use of the mathematical calculator and the early adoption of text synthesis and logical reasoning tools.

The calculator is introduced in the school environment as a tool for procedural offloading. It is made available only after the child has assimilated and internalized the basic operational concepts of arithmetic (addition, subtraction, multiplication, and division). At this stage, the student has already constructed the mental map of the operation, using the machine exclusively to automate mechanical calculation and accelerate processes. Prior mastery of the concept gives the student the reflective capacity to verify the obtained result—immediately identifying, for example, that if the display indicates 12 × 2 = 333, it is a typing error or system failure.

In impoverishing contrast, the early introduction of generative AI in the stages of reading and writing formation operates a premature cognitive offloading. The tool is activated even before the child has structured the mental schemas of syntax, textual coherence, and critical argumentation. By delegating to the machine the elaboration of essays, answers, and interpretations, the student skips the fundamental stage of reasoning gestation. Without the minimum conceptual repertoire to evaluate the validity of the generated text, the child becomes incapable of identifying hallucinations, factual errors, disinformation, or ideological biases produced by algorithms, compromising the formation of an autonomous and reflective mind.

4. DEVELOPMENT: CHATBOTS, EMOTIONAL DEPENDENCE, AND MENTAL HEALTH

4.1 The Chatbot as Confidant and "Virtual Friend"

Faced with deficits in interpersonal reception, socialization difficulties, or lack of access to specialized mental health services, a significant contingent of children and adolescents has begun using virtual companions and AI chatbots as affective confidants and informal therapists. The UNICEF data—revealing that more than 2 million minors actively seek personal advice on automated platforms—exposes the dimension of this vulnerability.

As analyzed by psychiatrist Dr. Daniel de Paula Oliva (Hospital Israelita Einstein, a Brazilian hospital), these systems operate through natural language processing models trained to mimic empathy, active listening, and affectionate vocabulary. However, artificial intelligence is completely devoid of consciousness, subjectivity, genuine empathy, and real capacity for care and relational attunement. The algorithmic simulation of reception creates an illusion of affective validation that contrasts with the deep and bidirectional nature of real human connections.

4.2 Psychological Consequences and Social Isolation

The anthropomorphization of conversational agents brings severe impacts to the psychic structuring of young audiences. The offer of a complacent artificial listening, totally predictable and immune to the conflicts inherent in real human relationships, induces a cycle of emotional dependence and social withdrawal.

Studies in the field of child psychiatry point to the following psychological consequences arising from this dynamic:

1. Replacement of Real Bonds: Progressive loss of interest in coexistence with peers, family members, and educators. The child begins to prefer aseptic digital interaction to the complexity of physical relationships, in which learning frustration and negotiation is indispensable.

2. Misleading Validation of Harmful Cognitions: Risks arising from the algorithm's ethical and technical inability to recognize and intervene in the face of self-destructive ideations, cognitive distortions, or emerging psychiatric disorders. By automatically "agreeing" with the user to retain their attention, AI can endorse dangerous thoughts.

3. Deepening of Loneliness and Isolation: Formation of shallow virtual connections that temporarily mask psychological suffering, but which, in the long term, aggravate the deep sense of helplessness, loneliness, and social disconnection.

5. DEVELOPMENT: DIGITAL SECURITY, CYBERCRIMES, AND PRIVACY

5.1 Increase and Sophistication of Digital Threats

The evolution of generative AI models has enhanced the scale, precision, and severity of cybercrimes directed at children and adolescents. According to alerts issued by the International Telecommunication Union (ITU/UN), by Cosmas Zavazava, and by UNICEF, algorithms have been instrumentalized by aggressors to automate and optimize exploitation strategies.

5.2 Improper Data Collection and Early Adultization

The functionality of artificial intelligence systems rests on the continuous and massive extraction of personal and behavioral data. In the case of child and youth audiences, this scraping encompasses sensitive records—including browsing history, voice patterns, emotional expressions, and geographic location—feeding databases used for commercial profiling and hyperspecific behavioral advertising targeting.

Child rights specialists emphasize that this unregulated commercial exploitation frontally violates the principles of comprehensive protection and absolute priority enshrined in Brazil's Child and Adolescent Statute (ECA). In addition to violating the fundamental right to privacy, the continuous display of content selected by engagement-oriented algorithms accelerates harmful processes of early adultization, exposing children to aesthetic standards, consumer appeals, and eroticized themes absolutely incompatible with their psychic maturity.

6. DEVELOPMENT: PEDAGOGICAL QUALITY, STEREOTYPES, AND BIASES

6.1 Automated Content and Reproduction of Prejudices

The emergence of children's productions generated entirely by AI—such as e-books, illustrated e-books, and synthetic animations produced on an industrial scale for rapid monetization on digital platforms, as reported by G1 Educação (a Brazilian news outlet)—has raised an alarm among educators about the degradation of the pedagogical quality of material consumed in childhood.

Analyses presented by professors Agnaldo Arraio (USP) and Débora Cardoso (Mackenzie, a Brazilian university) highlight the serious distortions present in these automated contents:

• Narrative Homogenization: Texts generated by artificial intelligence tend to replicate a shallow fable structure and a homogeneous and repetitive "moral of the story." As UNESCO warns, by depriving the child of contact with the plurality of worldviews and complex narratives, the expansion of their cultural and empathetic repertoire is limited.

• Euronormative Standardization and Under-representation: Image generators operate from markedly biased databases. Generic requests to illustrate child characters predominantly result in the creation of white, blonde, light-eyed human figures, excluding the representativeness of Black, Indigenous, non-Western-featured, or disabled children.

• Stimulation of Prejudiced Views: The passive assimilation of Euronormative content without diversity restricts children's racial and empathetic literacy, favoring the crystallization of discriminatory stereotypes. This lack of diversity later manifests in prejudiced behaviors and cyberbullying practices in the school environment.

7. DEVELOPMENT: GOVERNANCE, LEGISLATION, AND PROTECTION GUIDELINES

7.1 International Standards and Restrictions

The severity of potential harms associated with AI and digital platforms has driven a wave of legislative reviews globally. In late 2025, Australia approved pioneering legislation prohibiting access by minors under 16 to social networks, basing the decision on the understanding that systemic risks to child mental health outweigh the benefits declared by platforms.

In parallel, governments of nations such as the United Kingdom, France, Canada, and Malaysia are structuring new regulatory frameworks to impose strict duties of care and age restrictions on technology companies. At the institutional guidelines level, UNESCO and OpenAI itself establish the minimum age of 13 for the use of generative AI tools, although enforcement of this standard remains weakened by the absence of universal and secure age verification mechanisms.

7.2 The Regulatory Context in Brazil

In Brazil, the debate about regulating the digital environment and protecting minors' data has gained traction in the Legislative Branch and Federal Executive agencies. Bills under consideration in the National Congress aim to hold platforms accountable for the dissemination of illicit content, prohibit commercial collection of children's data, and criminalize cybercrimes enabled by artificial intelligence.

Within the Federal Government, the Ministry of Human Rights and Citizenship (MDHC) has issued official guides containing guidelines for child and youth digital security. Jurists and human rights advocacy entities reiterate that, although Brazil's Child and Adolescent Statute (ECA) predates the advent of AI, its doctrine of comprehensive protection and absolute priority has full legal effectiveness in cyberspace, binding the State, technology companies, and civil society in safeguarding minors' rights.

7.3 Priority Actions Recommended by UNICEF and the UN

To ensure that technological innovation does not proceed to the detriment of children's fundamental rights, UNICEF and UN agencies postulate five axes of priority and non-negotiable action:

1. Investment in Independent Scientific Research: Fund longitudinal and interdisciplinary studies dedicated to measuring the neurological, cognitive, and socioemotional impacts of AI use by children and adolescents.

2. Strengthening Criminal Legislation: Update legal frameworks to rigorously punish crimes of sexual exploitation, production of non-consensual deepfakes, and forms of grooming facilitated by intelligent systems.

3. Implementation of the Safety-by-Design Principle: Legally require technology companies to develop their products with native protective locks for child audiences, ensuring audit and transparency in algorithms.

4. Promote Digital and AI Literacy: Develop continuous educational programs for children, parents, teachers, and public policy makers, empowering society for critical and safe use of technology.

5. Reduction of Digital Inequalities with Security: Expand infrastructure and access to learning technologies ensuring that digital inclusion occurs in properly protected and equitable virtual environments.

8. CONCLUSION

8.1 Integrative Synthesis

The convergence of scientific evidence, demographic indicators of use, and alerts issued by specialists and multilateral entities demonstrates unequivocally that the risks of artificial intelligence for childhood do not constitute isolated events, but rather a systemic and multidimensional threat:

                          ┌──────────────────────────────────────────┐
                          │    MULTIDIMENSIONAL IMPACTS OF AI        │
                          └────────────────────┬─────────────────────┘
                                               │
        ┌──────────────────────┬───────────────┴───────────────┬──────────────────────┐
        │                      │                               │                      │
┌───────▼────────┐    ┌────────▼─────────┐           ┌─────────▼─────────┐    ┌───────▼────────┐
│  COGNITIVE     │    │  MENTAL HEALTH & │           │ DIGITAL SECURITY  │    │  PEDAGOGICAL   │
│  DEVELOPMENT   │    │  AFFECTIVITY     │           │   & PRIVACY       │    │   QUALITY      │
├────────────────┤    ├──────────────────┤           ├───────────────────┤    ├────────────────┤
│ Loss of focus, │    │ Chatbot depend-  │           │ Grooming, sexual  │    │ Race and gender│
│ overload, and  │    │ ence, false      │           │ deepfakes, and    │    │ biases, homo-  │
│ critical rust- │    │ empathy and      │           │ illegal data      │    │ geneous morals │
│ ing.           │    │ social isolation.│           │ collection.       │    │ and prejudices.│
└────────────────┘    └──────────────────┘           └───────────────────┘    └────────────────┘

Early and unmediated immersion in unregulated AI ecosystems compromises the formation of analytical reasoning, replaces the affective bonds indispensable for psychological maturation with simulated interactions, exposes children to sophisticated forms of digital violence, and impoverishes the worldview of new generations through algorithmic filters and biases.

8.2 Call for Transversal Action

As emphasized in UNICEF and UN guidelines, the international community finds itself at a historic crossroads. The regulatory, legal, and ethical decisions adopted in the present regarding the governance of artificial intelligence will seal the conditions of security, privacy, mental health, and full development of children in the coming decades.

Protecting childhood in the age of artificial intelligence requires a pact of shared responsibility: the State must enact firm laws and exercise rigorous oversight; technology companies must assume the ethical and legal duty to deliver safe products by design; educational institutions must promote critical digital literacy; and family units must exercise attentive mediation in minors' daily lives. Technological advancement will only be genuinely evolutionary if it safeguards, first and foremost, the integrity and future of its children.