UNICEF Issues 10-Point AI Guidelines to Protect Children Ahead of Global Delhi Summit

On the eve of the Artificial Intelligence (AI) summit in New Delhi later this month, UNICEF has issued a detailed guideline about the "opportunities and risks to children, young people and their families, teachers and communities." The UNICEF paper speaks of the Need to optimize opportunities, mitigate risks and eliminate harms, including child abuse and exploitation for children." UNICEF, while addressing governments, business and other organisations, speaks of the need to gather enough evidence on how "AI impacts children and their environments, ideally through participatory research and decision making and to centre children in AI policies." UNICEF's 10-point recommendations are:* ensuring a regulatory framework, oversight and compliance

for child-centred AI. This would include continuously assessing and monitoring AI's impact on chidlren throughout the entire AI development life cycle amnd "prevent AI-enabled crimes against children such as child sexual abuse and exploitation." UNICEF said AI-enabled chatbots can be risky and "supervised safety training" is necessary.*Governments should have rules that protect children from the harmful use of AI in armed conflict and cyber operations. * As children are vulnerable, their data including biometric data needs to be protected and governments and businesses should address children's privacy in AI policies..* To ensure "non-discrimination," it is important to support "children in disadvantaged or vulnerable situations so that they benefit from AI systems." There should also be no prejudicial bias against children or groups of children.* As for transparency, "explainability" and accountability, UNICEF said that it was "important to avoid a situation where children and their caregivers think they are interacting with a human. AI systems should immediately and transparently provide a warning before starting an interacting, explaining in a clear and forthright manner that they are about to start interacting directly or indirectly with an AI system." Age appropriate language is also necessary.* Respect for children's rights is essential as is the need to eliminate child labour and exploitation. * It is important, said UNICEF, to keep the best interests of children in mind, including their development and well-being. * AI should also be for children and the participation of children, keeping in mind policy and governance and also, design and development must be kept in mind. Apart from recommending an "inclusive design approach," there is a need for diversity among those who work with AI, including designing and developing AI systems. * If children have to learn to use AI, there must be formal and informal educational programmes for them and teachers too have to be trained and guided so that they can perform well and eliminate risks. * Finally, there is a realisation that a digital divide exists in the world and it is necessary to support infrastructure development to ensure equitable sharing of AI products. Besides, creating child-centre AI policies and providing incentives for them and working on reaching all children is imperative. Naturally, an equal opportunity AI for children works best.