Guidance on AI and children
Version 3.0 | Recommendations for AI policies and systems that uphold child rights
Highlights
As part of our AI for children project, UNICEF has updated its guidance (version 3) to promote children's rights in government and private sector AI policies and practices, and to raise awareness of how AI systems can uphold or undermine these rights.
Drawing on the Convention on the Rights of the Child, the guidance offers 10 requirements for child-centered AI:
- Ensure regulatory frameworks, oversight and compliance for child-centred AI
- Ensure safety for children
- Protect children's data and privacy
- Ensure non-discrimination and fairness for children
- Provide transparency, explainability and accountability for children
- Respect human and child rights through responsible AI practice
- Support children's best interests, development and well-being
- Ensure inclusion of and for children
- Prepare and skill children for present and future developments in AI
Create an enabling environment for child-centred AI
A checklist of the requirements and recommendations is also available.
Why now?
This third edition of the UNICEF guidance on AI and children was prompted by rapid advances in AI technologies, such as generative AI, increased adoption of AI systems by children, emergent opportunities and risks with regard to children’s rights, and changes in the AI governance landscape.
The guidance has been updated throughout the document, where needed, with noteworthy additions focusing on: AI companions used by children; the AI supply chain, including child labour and datasets contaminated with harmful and illegal content; the environmental impacts of AI systems, which particularly impact children; AI-generated child sexual abuse material (CSAM) and non-consensual intimate images (NCII); and AI in armed conflict and cyber operations. Opportunities to seize include the use of AI systems to better support learning and increase accessibility for children with disabilities, as well as leveraging AI to improve child well-being.
While AI is increasingly a part of children’s lives, guidance on how AI policies and systems can uphold children’s rights remain scarce. The significant changes in the AI landscape, and this gap, prompted the update.
Watch the video of the launch webinar here:
How the guidance was developed
The draft policy guidance (version 1.0) was released in September 2020. UNICEF subsequently sought input from stakeholders working in areas related to the fields of AI and children’s rights. This included AI developers and deployers, companies, government agencies, civil society, international organizations, academics and adult and child citizens. We invited stakeholders to express their views on the draft guidance and provide feedback and comments (see privacy notice.) Click here to find out how they responded to the draft guidance.
This input was analysed and incorporated into version 2.0 (2021) (English | Français | Español | عربى). In order for policy guidance to address implementation complexities, it needs to be put to use by policymakers, public organizations and businesses for validation and local adaptation. We thus invited governments and the business sector to pilot the draft guidance in their field and openly share their findings about how it was used, and what worked and what did not, so that their real experiences could improve the document. In this spirit, UNICEF worked with a group of government and business “pilot partners” to develop case studies on how each will implement the guidance: Read the case studies. Piloting organizations adhered to these terms.
To support implementation of version 2, a set of practical implementation tools were developed in 2021: AI guide for parents (English | Français | Español | عربى), AI guide for teens (English | Français | Español | عربى), and a Roadmap for policymakers and AI for children development canvas (English | Français | Español | عربى).
Version 3 (2025) was informed by a range of inputs: the guidance of a diverse expert advisory group; an expert consultation with civil society, governments and academia; internal and external peer review; and emergent evidence, including a twelve-country UNICEF study with children and caregivers. This consultative and broad approach ensured the recommendations reflect real-world experiences across regions.
