WEF expert calls for equal focus on AI innovation and child safety

World Economic Forum's Cathy Li called for integrating AI literacy into India's education curriculum

With AI becoming part of everyday life, a top global expert on artificial intelligence has called for integrating AI literacy into India's education curriculum and creating incentives for AI companies to invest in child safety.

Noting that opportunities and risks presented by AI are global, Cathy Li, Head of the Centre for AI Excellence at the World Economic Forum, emphasised that solutions must be localised, especially for India given its diversity, and safeguards need to work across languages and cultures.

Li, co-author of a whitepaper titled 'Children's Safety and Well-Being in the Age of AI: Best Practices and Incentives', told PTI in an interview that AI systems in India need to be tested across languages and cultural contexts, but also for different levels of connectivity, digital literacy and device access.

The recently released report said AI can unlock major benefits for children, but only if safety, agency, inclusion, and accountability are built in.

Asked about specific recommendations for India for developing an AI-ready workforce by pairing AI literacy with child-safe design, Li said AI literacy first needs to be made part of education.

"Children need to understand where AI can fail, how to verify information and when human judgement is needed. Second, build safety and child-rights assessments into systems likely to be used by children," she said.

She also called for testing AI models across India's linguistic and cultural diversity, involving children, educators, caregivers and child-safety experts.

Fourth, use procurement, funding and independent assurance to create incentives for companies to invest in child safety.

"And finally, don't treat safety and innovation as competing goals. Children are going to grow up with AI. The objective should be to give them the skills and protection to use it well," she added.

Noting that AI is becoming part of how children learn, create, communicate, play and access support, the WEF whitepaper argued that AI can expand education, accessibility, language access and creativity.

"Yet generative, conversational and agentic systems can also generate, personalise, persuade and act in ways that intensify harm. Children should not have to surrender privacy, agency, human connection or equitable access to participate in an AI-enabled world," it said.

Asked about specific opportunities and risks that AI presents for children in India, Li said the opportunities and risks are global, but how they play out is local.

"AI can expand access to education, creativity and language support, but it can also expose children to harmful content, manipulation, privacy risks and exploitation," she noted.

Li said India's scale and diversity are particularly important and safeguards need to work across languages, cultures, levels of connectivity and very different family and educational settings.

"We cannot design only for a child with reliable internet access, an English-language interface and a digitally literate household. So, we need a common baseline for child safety, but it has to work in the context in which children actually use AI," she added.

Underlining that there is no single safeguard, Li said protection needs to reflect the child's age and development, but also what the AI system can actually do. An age gate alone does not make a system safe.

"Second, AI should support learning and independent judgement, not replace teachers, caregivers or human relationships.

"And third, children need AI literacy. They need to know that AI can be wrong, how to question what it tells them and when to turn to a person for help," she said.

Li said a safeguard that works in one language but fails in another is not an effective safeguard.

She further called for keeping the safety tools simple.

"Privacy settings, parental controls and reporting mechanisms should be easy to use, with strong protection on by default. A child should not be less protected because their parents have less time, fewer resources or less digital experience.

"And children, parents and educators should be part of testing these systems. They will see things that developers may not," she said.

Li further said a young child and a 17-year-old do not have the same needs, and neither do two children of the same age in very different circumstances.

"But we also shouldn't have to reinvent child safety country by country. There should be a common baseline around privacy, protection from exploitation, accountability and human oversight, with implementation adapted to local realities.

"And the use case matters. A homework tool presents very different risks from an AI companion or an agent that can take actions on a child's behalf. The safeguards should reflect that," she said.

Li said India has an opportunity to build this capacity into its AI ecosystem as it grows and it can test AI in Indian languages and contexts, with local data sets, independent evaluation, reporting mechanisms and child-rights impact assessments.

"There is also a role for shared infrastructure. A start-up cannot necessarily build the same testing capacity as a major technology company. Government, industry, universities and civil society can help provide shared tools, data sets and expertise.

"Ultimately, India should have the capacity to assess whether systems are safe for children in India, rather than relying only on testing conducted elsewhere," she said.

"For India, I think the lesson is to combine common principles with local implementation. Expectations around privacy, accountability, human oversight and protection from exploitation should be clear. But how those protections work needs to reflect India's languages, education systems, connectivity and patterns of technology use.

"We should not measure success simply by whether children have been kept away from AI. We should ask whether they are safer and better equipped to use it," she said.