05th May 2026
1. Executive Summary
Amnesty International Kenya, and Watoto Watch Network welcome the opportunity to submit our views on the Artificial Intelligence Bill, 2026 (“the Bill”), introduced into the Senate on 2nd April 2026 by Senator Karen Nyamu. This submission addresses a single, critical gap in the Bill: the complete absence of any provision protecting children from the specific harms that artificial intelligence systems pose.
We affirm that child protection in the context of artificial intelligence requires deliberate consideration of children’s unique vulnerabilities and the provision of appropriate safeguards necessary for their wellbeing. This recognition dispels the assumption that children’s rights and protections are sufficiently secured under general provisions alone. We therefore maintain that children should be accorded special status within the Bill, with specific and enhanced protections beyond those applicable to the general public.
This special status is recognized under the Constitution of Kenya, the Children Act, 2022 (No. 29 of 2022), the United Nations Convention on the Rights of the Child, the African Charter on the Rights and Welfare of the Child and even the European Union Artificial Intelligence Act (Regulation (EU) 2024/1689). It is grounded in the recognition that children are inherently vulnerable and therefore require a heightened level of legal and regulatory protection. Failure to recognize this special status exposes children to heightened AI-related harms by subjecting them to the same legal and regulatory standards as adults, despite their evolving capacities and limited ability to meaningfully safeguard their own interests within digital environments. Such an approach undermines both constitutional and international child protection obligations.
This memorandum identifies four legislative gaps and makes recommendations anchored in comparative provisions of the EU AI Act and the applicable standards domestically under the Children Act 2022 having article 53 of the Constitution of Kenya as the overarching anchor.
2. The Children Act, 2022: The Applicable Domestic Standard
The Children Act, 2022, is Kenya’s primary statute governing children’s rights. It gives effect to Article 53 of the Constitution and implements Kenya’s obligations under the UNCRC and the ACRWC. The following provisions of the Children Act, 2022 are directly relevant to the identified gaps in the AI Bill:
| Provision | Content and Relevance |
| Section 2 | Defines “child” as “an individual who has not attained the age of eighteen years.” Also defines “child abuse” to expressly include: (h) the use or exposure of a child in electronic or online platforms for purposes of pornography or any other unlawful sexual practice; and (i) the use of a child’s images for purposes of pornography or sexual gratification. The definition of “grooming” under Section 2 explicitly covers electronic means. The AI Bill contains no corresponding definition of a child which inadvertently denies children the recognition of being a vulnerable group which then exposes them to AI-related risks. |
| Section 3 (Objectives) | The Act exists to give effect to the Constitution on matters relating to children. The AI Bill, which will govern AI systems deployed in schools and other child-accessible environments, must bear and consider the best interests of the child by design. |
| Section 22(1) and (3) | Prohibits psychological abuse and child abuse. Section 22(3) creates specific offences for subjecting a child, through any electronic system or network, to: (a) solicitation for sexual purposes; (b) transmission of obscene material; and (c) online abuse, harassment or exploitation; including cyber-bullying, grooming, and cyber-stalking. AI-generated content and recommendation systems capable of facilitating these harms fall squarely within this section, yet the AI Bill has no specific provisions to mitigate these harms. |
3. The EU AI Act: The Comparative Standard
The Bill’s Memorandum of Objects and Reasons explicitly states that it aims to align with “global standards including the European Union Artificial Intelligence Act.” Regulation (EU) 2024/1689 (the EU AI Act), which entered into force on 1 August 2024, contains several provisions specifically addressing children that are absent from Kenya’s Bill. The following are the directly relevant provisions:
| Provision | Content and Relevance |
| Article 2(9) and (11) Recital 11 (PDF p.11) | Preserves national laws protecting minors (persons below 18), “taking into account the United Nations General Comment No 25 (2021) on children’s rights in relation to the digital environment.” The EU AI Act identifies the protection of minors in the digital environment specifically within AI systems. Kenya’s AI Bill has no such recognition of minors and their protection in AI systems. |
| Article 5(1)(a) Recital 28 (PDF p.29) | Recognizes “the rights of the child” as those that could potentially be abused by Artificial Intelligence which could be misused and provide novel and powerful tools for manipulative, exploitative and social control. The Act goes ahead to prohibit such abuse. Kenya’s Bill doesn’t accord such a recognition to the dangers that AI poses to children’s rights, not to mention prohibiting practices that would endanger such rights. |
| Article 9(9) Recital 48 (PDF p.44) | Recognizes the adverse impact caused by AI systems to fundamental rights when classifying AI systems as high risk. The Act goes further and gives recognition to children’s specific rights as enshrined in Article 24 of the Charter and in the United Nations Convention on the Rights of the Child, further developed in the UNCRC General Comment No 25 as regards the digital environment, both of which require consideration of the children’s vulnerabilities and provision of such protection and care as necessary for their wellbeing. Children are as such granted a high level of protection in the digital environment in assessing the severity of the harm that an AI system can cause. Kenya’s AI Bill contains no equivalent recognition off children as vulnerable and doesn’t consider children’s vulnerabilities in assessing AI systems. |
| Article 5(1)(b) (PDF p.182) | Prohibits AI systems that exploit “any of the vulnerabilities of a person or a specific group of persons due to their age, disability or a specific social or economic situation, with the objective, or the effect, of materially distorting the behavior of that person […] in a manner that causes or is reasonably likely to cause that person […] significant harm.” Kenya’s Bill has no equivalent prohibitions on vulnerabilities of persons due to their age, in this case, children. |
| Article 7(2)(h) (PDF p.195) | Requires that “the extent to which there is an imbalance of power, or the persons who are potentially harmed […] are in a vulnerable position […] in particular due to status, authority, knowledge, economic or social circumstances, or age” be a factor in high-risk classification. The proposed Bill contains no equivalent age-vulnerability criterion, not to mention a definition of what a vulnerable position is. |
4. Identified Gaps and Proposed Amendments
The following four gaps are identified through systematic cross-reading of the AI Bill against the Children Act, 2022, the EU AI Act, and Kenya’s international obligations. Each gap is presented with the specific AI Bill sections engaged, what those sections already do well, the exact proposed amendment wording, and the reason for the amendment.
1). The Law Must Name Children
| Bill Section / Clause | Provision Commended or Engaged | Proposed Amendment (Exact Wording) | Reason for Proposed Amendment |
| Section 2 Interpretation Section 3 Objects of the Act | Section 2 provides a comprehensive set of definitions covering AI systems, providers, deployers, and users. Section 3 lists the protection of human rights and data protection as objects of the Act. | Insert the following definitions in Section 2: “child” means a person below the age of eighteen years, consistent with the definition under Section 2 of the Children Act, 2022 (No. 29 of 2022); “child-directed artificial intelligence system” means an artificial intelligence system that is: (a) designed or marketed for use by children; (b) reasonably foreseeable to be accessed by children as a significant portion of its user base; or (c) deployed in an educational, health, or social care setting serving children. | The Bill defines users, deployers, providers, and AI systems but says nothing about children. A law cannot protect those it does not name. The Children Act, 2022, Section 2 defines a child clearly as anyone under 18. The EU AI Act, Recital 11 uses the same definition. The AI Bill must be consistent with Kenya’s own domestic law. Without this definition, children are treated identically to adults throughout the Bill, with no recognition of their heightened vulnerability, reduced legal capacity, or the specific rights they hold under Article 53 of the Constitution and the Children Act, 2022. |
2). Children Deserve Their Own Protections, Not Borrowed Adult Rules
| Bill Section / Clause | Provision Commended or Engaged | Proposed Amendment (Exact Wording) | Reason for Proposed Amendment |
| Section 26 Obligations for high-risk AI systems Section 28 Transparency and safeguards Section 30 Ethical guidelines Section 31 AI literacy | Section 26(1)(b) requires human rights impact assessments before deployment of high-risk systems. Section 26(1)(g) requires explicit consent before generating or manipulating images or likeness. Section 28 requires transparency disclosures to users. Section 31 provides for AI literacy programmes in partnership with educational bodies. | Insert a new Section 26A titled “Additional obligations for child-directed and child-accessible artificial intelligence systems”: (1) A provider or deployer of a child-directed artificial intelligence system shall, in addition to the obligations under Section 26: (a) ensure that the system meets age-appropriate design standards, including disclosures in plain language suited to the developmental stage of child users; (b) obtain verifiable parental or guardian consent before processing personal data of a child user, consistent with Section 27(3) and Section 8(1) of the Children Act, 2022; (c) not deploy behavioural advertising, commercial profiling, or recommendation algorithms targeting child users for commercial purposes; (d) not incorporate emotion recognition, attention tracking, or real-time biometric identification of children, consistent with Article 5(1)(f) of EU AI Act Regulation (EU) 2024/1689; (e) not deploy AI systems that exploit age-based vulnerabilities to distort the behaviour of child users in a harmful manner, consistent with Article 5(1)(b) of EU AI Act Regulation (EU) 2024/1689; (f) conduct a Child Rights Impact Assessment, in addition to the human rights impact assessment required under Section 26(1)(b), before deployment of any system likely to be accessed by children. (2) The Commissioner shall develop guidelines for child-appropriate AI design in consultation with the Cabinet Secretary responsible for children’s affairs, the Data Protection Commissioner, and civil society organisations with expertise in children’s rights and digital rights. | The rights of a child are distinct rights, not a smaller version of adult rights. They exist precisely because children are at a different stage of development, have less legal capacity, and are more vulnerable to manipulation and exploitation. Section 28’s transparency disclosures are designed for cognitively mature adult users. A disclosure that satisfies an adult may be wholly meaningless to a child of eight or twelve years. Section 26(1)(g)’s reference to “legal representative” implicitly captures children but does not require verifiable parental consent or impose any additional child-specific safeguard. Children’s attention and behaviour must not be commercial resources. Children’s emotions must not be monitored by AI systems in schools where they have no meaningful choice about participation. Evidence from Amnesty International’s own research confirms the urgency of these protections. In “Driven into Darkness: How TikTok’s ‘For You’ Feed Encourages Self-Harm and Suicidal Ideation” Amnesty International documented how algorithmic recommendation systems; the same class of AI systems that the Bill’s Section 26 is designed to regulate, actively push children into escalating cycles of harmful content. In the Kenyan manual experiment, once the “rabbit hole” effect took hold at the 20-minute mark, 72% of videos recommended in the following 40 minutes related to mental health struggles, with at least five references to suicidal thinking or death wishes. Not a single mental health-related video in that sequence was produced by a mental health professional or recognized organization. Children and young people in Kenya, the same jurisdiction this Bill will govern, spend just over three hours per day on social media, one of the highest rates globally, making them disproportionately exposed to AI-driven content recommendation systems. TikTok’s algorithmic system was found to exploit watch time and engagement data to infer children’s emotional state and mental health interests, then use those inferences to serve progressively harmful content, without the child’s knowledge or meaningful consent. This is precisely the harm that the proposed Section 26A(1)(c) and (e) are designed to prevent: recommendation algorithms that exploit children’s developmental vulnerabilities and distort their behaviour in harmful ways. The Bill’s current provisions, requiring only generic transparency disclosures under Section 28 ,would have done nothing to prevent the harms documented in Kenya. Basis: EU AI Act Articles 5(1)(b) and (f), Recitals 28, 29, 48; Children Act, 2022, Sections 8, 22, and 27; UNCRC General Comment No. 25 (2021); Amnesty International, “Driven into Darkness: How TikTok’s ‘For You’ Feed Encourages Self-Harm and Suicidal Ideation”, November 2023. |
3). When Children Are Involved, the Risk Is Always Higher
| Bill Section / Clause | Provision Commended or Engaged | Proposed Amendment (Exact Wording) | Reason for Proposed Amendment |
| Section 25 Classification of AI systems Section 25(2)(b) High-risk- education sector | Section 25(2)(b) correctly classifies AI systems used in education as high-risk. This is a sound and welcome provision. Section 25(1) classifies systems by the risk they pose to health, safety, fundamental rights, the environment, or societal welfare. | Amend Section 25 to insert a new subsection (2A): “(2A) In classifying artificial intelligence systems under this section, the Commissioner shall have particular regard to: (a) whether the system is used by, marketed to, or likely to be accessed by children; and (b) whether the system has the potential to exploit the vulnerabilities of children due to their age, developmental stage, or reduced legal capacity; and shall, in all such cases, apply a presumption in favour of a higher risk classification.” | The classification framework contains a critical blind spot: it does not ask whether the people affected are children. A system that poses a moderate risk to an adult may pose a severe risk to a child. A system that an adult can critically evaluate and reject may be one that a child absorbs without question. Basis: EU AI Act Article 7(2)(h) (age and vulnerability as high-risk factors); EU AI Act Recital 48; Children Act, 2022, Section 8(1). |
4). The Ethics of AI Must Speak Directly to the Protection of Children
| Bill Section / Clause | Provision Commended or Engaged | Proposed Amendment (Exact Wording) | Reason for Proposed Amendment |
| Section 30 Ethical guidelines Section 30(2)(a) Vulnerable groups | Section 30(1) requires the Commissioner to develop and publish ethical guidelines for AI development, deployment, and use. Section 30(2)(a) requires that those guidelines address prevention of bias and exclusion “with particular regard to vulnerable groups.” Section 30(3) requires the guidelines to be developed in consultation with relevant agencies and stakeholders. | Amend Section 30(2) by inserting a new sub-paragraph (g): “(g) the specific vulnerabilities and rights of children in relation to artificial intelligence systems, including: (i) protection from manipulation and exploitation of developmental vulnerabilities by AI systems; (ii) protection from commercial profiling of children by AI systems; (iii) age-appropriate design requirements for AI systems accessible to children; and (iv) governance of AI systems used in educational settings, ensuring they serve children’s education rather than commercial interests; with guidelines to be developed in consultation with the Cabinet Secretary responsible for children’s affairs and the National Council for Children’s Services established under Section 41 of the Children Act, 2022, and taking into account the Children Act, 2022, UNCRC General Comment No. 25 (2021), and Article 5(1)(f) of EU AI Act Regulation (EU) 2024/1689.” | The ethical guidelines are one of the most powerful provisions in the Bill because they shape how AI systems are built and evaluated long before any harm occurs. But Section 30(2)(a)’s reference to “vulnerable groups” does not name children, creates no specific content obligation, and is therefore unenforceable in practice. Children’s rights cannot be an afterthought in ethical guidelines that will define how AI treats every person in Kenya. The EU AI Act recognizes the rights of the child as a special category in Recital 28. Kenya’s guidelines must do the same. Basis: EU AI Act Recital 28; Children Act, 2022, Sections 22 and 27; UNCRC General Comment No. 25 (2021). |
5. Conclusion
The Artificial Intelligence Bill, 2026 is a landmark and commendable piece of legislation. The establishment of an independent Artificial Intelligence Commissioner, a four-tier risk classification framework, transparency obligations for providers and deployers, and regulatory sandboxes for innovation are all welcome and necessary developments.
First, the Bill does not define a child anywhere in its text and does not recognize children as a category requiring specific protection under Section 2 or Section 3. Second, the Bill’s transparency and safeguard obligations under Sections 26, 28, 30, and 31 are designed for adult users and impose no age-appropriate design requirements, no obligation to obtain verifiable parental consent, and no prohibition on commercial profiling of children or the deployment of emotion recognition systems against child users. Third, while Section 25(2)(b) correctly classifies education as a high-risk sector, the Bill does not require the AI Commissioner to treat the involvement of children as an aggravating risk factor in classification decisions, and does not explicitly name attendance monitoring, behavioural analysis, or content recommendation in schools as high-risk applications. Fourth, Section 30’s ethical guidelines framework does not mandate a single word about children, the reference to “vulnerable groups” in Section 30(2)(a) is vague. Each of these gaps has a targeted remedy in this submission: a definition of “child” and “child-directed AI system” in Section 2; a new Section 26A establishing child-specific obligations for providers and deployers; an amendment to Section 25 requiring the Commissioner to presume higher risk where children are involved; and an amendment to Section 30(2) mandating child-specific ethical guidelines. Taken together, these amendments would bring the Bill into alignment with Kenya’s obligations under the Children Act, 2022, Article 53 of the Constitution, the UNCRC, and the EU AI Act; the very framework the Bill’s own Memorandum commits to following.
Parliament is respectfully urged to:
- Refer this memorandum to the relevant Senate Standing Committee for consideration during the Second Reading and Committee Stage of the Bill; and
- Ensure that any amendments at Committee Stage include, at minimum, a definition of “child,” explicit recognition of children as a protected category, a prohibition on AI systems that exploit age-based vulnerabilities (equivalent to EU AI Act Article 5(1)(b)), and a prohibition on emotion recognition AI in educational institutions (equivalent to EU AI Act Article 5(1)(f)).
Submitted for parliamentary consideration.
Amnesty International Kenya Watoto Watch Network