AI与儿童指南3.0

Guidance on AIand Children Updated guidance for governments andbusinesses to create AI policies and systemsthat uphold children’s rights Contents Introduction1 Foundations for child-centred AI Requirements for child-centred AI 1.Ensure regulatory frameworks, oversight and compliance for child-centred AI2.Ensure safety for children3.Protect children's data and privacy4.Ensure non-discrimination and fairness for children5.Provide transparency, explainability and accountability for children6.Respect human and child rights through responsible AI practice7.Support children's best interests, development and well-being8.Ensure inclusion of and for children Implementation Endnotes Acknowledgements Introduction Collectively, these changes bring opportunities and risks to children, youngpeople and their families, teachers and communities. New benefits thatcould be leveraged include the use of AI systems to better support learningand increase accessibility for children with disabilities. Novel risks includeAI-generated disinformation and emotional dependency on companion 37% of Americanelementary Since 2021, there has been a much-needed uptick in efforts to support andprotect children in an AI world. From a governance perspective, these include37% of childrenaged 9–11 inArgentina turn child-related legal stipulations in the EU AI Act4and the Council of EuropeFramework Convention on AI and human rights, democracy and the rule oflaw.5Recommendations appear in the Joint Statement on AI and the Rights Turing Institute, the Children’s Parliament and the Scottish AI Alliance,9as wellas by research groups like Digital Futures for Children with EU Kids Online.10 Yet, despite children being at the forefront of AI adoption, little isknown about the mid- to long-term impact of AI on them; for example,developmental (cognitive and psychological) and learning impacts, as wellas impacts on the societies in which they live. In addition, children remain atthe margin of shaping AI systems. This is especially true for children from theGlobal South, for whom location, digital divides and severely limited access AI policies andsystems shouldprotect children,provide equitablyfor their needs andrights and support The notable policy, research and engagement efforts towards child-centredAI listed above are the exception, not the norm. Children’s rights are still notreceiving sufficient attention in AI policy, law, governance and development.11As noted, there is very limited understanding of how this unprecedented The message is clear: in order to uphold children’s rights, AI governanceand systems need to optimize opportunities, mitigate risks and eliminateharms for children. The need to address the evidence gap on how AI impacts This guidance draws on the UN Convention on the Rights of the Child(CRC)12to lay out the foundations for child-centred AI: today and in thefuture, AI policies and systems should protect children, provide equitably Children should be empowered with access to, and opportunity to benefitfrom, AI systems. Building on this foundation, the guidance presentsten requirements for child-centred AI, complementing key work already The guidance is accompanied by many resources, including previous versions,eight implementation case studies, guides for parents and teens, opinion The purpose of the guidance: •Raise awareness of children’s rights and how AI systems can uphold orundermine those rights; and Target audience: Since most AI policies are devised by government actors, and most AIsystems are developed, procured, designed or implemented by governments,State actors and those in the business sector, we have continued to primarily •Government stakeholders: Including all governance actors at theregional, national or local level who create AI strategies, policies andlaws – such as policymakers and legislators – and those responsible for •The business sector: Businesses that procure, design, develop, deploy oruse AI systems, including functions involved in creating policies, guidesand codes of conduct for their companies, and teams that implement In addition, this guidance will hopefully add value to the efforts of otherstakeholders in the AI governance and implementation ecosystem, includingUN bodies, civil society organizations, non-governmental standardsdevelopment organizations and academia, all of which are involved in better The guidance can be used in a variety of contexts: •When creating, reviewing and/or updating AI guidelines, codesof conduct, strategies, policies, laws or regulations, or industry/technical standards;•When designing, developing and deploying AI systems that childreninteract with or may be impacted by (even if not designed to be usedby children), which includes consideration of the content, decisions andinteractions those systems produce;•When procuring and deploying externally developed systems, whichis an increasingly common practice and entails responsibilities andrequireme