In a quiet British nursery, a baby who has not yet learned to speak reaches not for a parent’s arm but for the glowing rectangle on the blanket. The gesture, documented in a landmark review of global research by four UK universities, is not anecdotal but a pattern: researchers found that some infants are increasingly turning to digital devices for comfort instead of seeking reassurance from a caregiver. The review stopped short of establishing direct causation between screen use and developmental harm, yet the accumulation of evidence — delayed language, disturbed sleep, overstimulation, a retreat from physical play — was, in the researchers’ words, compelling enough to demand clearer guidance. Andrea Leadsom, founder of the 1,001 Critical Days Foundation, called the findings “a wake-up call” for the period between pregnancy and age two, the most consequential stretch of human development.

That same week, in a conference hall in Geneva, the United Nations secretary-general opened the first Global Dialogue on AI Governance with a warning that no child should become “a guinea pig of an unregulated AI”. António Guterres proposed a global pact built on three principles: companies must prove their systems are safe for minors before deployment; there must be zero tolerance for AI-generated child sexual abuse imagery; and platforms should be obliged to refer children to human help when signals of distress or self-harm are detected. The urgency was underlined by UNICEF data showing that at least 20 million children in ten countries have already used AI tools, adopting them at more than three times the rate of adults. Over two million have turned to these systems for personal advice, often treating them as neutral, trustworthy confidants without understanding how their data is handled or what values are baked into the responses.

Viewed from Washington, the public mood is darkening. Pew Research Center figures show that the share of American adults more concerned than excited about AI rose from 37 per cent in 2021 to 52 per cent two years later, with a majority now expecting negative effects on both personal life and society. Analysts in London note that this erosion of trust mirrors earlier trajectories: genetically modified foods and childhood vaccines both saw public confidence collapse not because of new scientific evidence but because of a perceived failure of institutions to manage risk transparently. In the Gulf, the conversation is shifting from parental responsibility to infrastructure. With internet penetration in the UAE at 99 per cent and one in three users globally a child, telecom operators and regulators are being urged to embed safety at the network level, automatically blocking known abuse material before it ever reaches a screen. The Internet Watch Foundation’s 2025 report recorded a 26,385 per cent year-on-year increase in AI-generated child sexual abuse videos, a statistic that has made network-level filtering, in the words of one regional analyst, “not an optional extra but a baseline expectation”.

From Jakarta, meanwhile, came a different inflection of the same anxiety. At a G20 Sherpa meeting in Washington, Indonesian officials insisted that global AI standards must not become a new compliance barrier that discriminates against micro, small and medium enterprises in developing countries. The demand was for flexibility and co-authorship: “Developing countries must be co-authors in the formulation of standards, not merely implementers of compliance,” said deputy minister Edi Prio Pambudi. Indonesia is simultaneously drafting its own national AI roadmap, pushing for cross-border QR payment interconnections to lower costs for migrant workers, and backing a ban on using pirated content to train models — all while insisting that international rules respect national data protection laws. The image that lingers is not of a conference room but of a child’s hand, suspended between a parent’s touch and a screen’s glow, while far away in windowless data centres, the same technology that streams nursery rhymes is being rewired, slowly and imperfectly, to act as a silent guardian.