UN calls for children to be consulted on AI before decisions are made, as UNICEF counts 1.2 million victims of fake images

By UN News · September 22, 2026.

More than 1,000 children from 49 countries have put a concrete request to the UN: they want a say on artificial intelligence (AI) before adults decide for them, not afterwards. They delivered it at a high-level meeting held on Tuesday in New York, alongside the General Assembly's general debate, co-chaired by French President Emmanuel Macron, Spanish Prime Minister Pedro Sánchez, Kenyan President William Ruto and Australian Prime Minister Anthony Albanese. "Today's children are the first generation to grow up with artificial intelligence, but often the last to be asked their opinion," said Volker Türk, UN High Commissioner for Human Rights.

The figure behind the urgency comes from UNICEF: according to its research, at least 1.2 million children in 11 countries saw their image turned into fake sexual content generated with AI within the space of a single year. "Despite its name, there is nothing fake about the harm this content causes children," warned Catherine Russell, UNICEF's executive director. It is a figure that leaves no room for ambiguity: it speaks not of a hypothetical future risk but of harm that is already massive and measurable, although the source does not specify the exact period that year covers.

The consultation driven by Türk's office, which surveyed more than 1,000 minors in 49 countries, also reveals what worries children themselves when they think about AI: the creation of fake images of real people, the exploitation of their personal data, a possible loss of creativity and the risk that these tools replace relationships with their family, their friends, their teachers or their therapists. It is a remarkably mature catalogue of fears, and it largely matches what already worries adult psychologists and educators.

The meeting falls under a coalition launched in July in Geneva, the Coalition for Children's Rights and Protection in the Age of Artificial Intelligence, which brings together governments, UN agencies, technology companies, civil society and experts, and which 17 countries have joined so far. Türk is asking for something more ambitious than good intentions: that the Convention on the Rights of the Child —the most widely ratified human rights treaty in the world— also apply to AI, with risk assessments before products are launched on the market, greater transparency and independent oversight. His sharpest line was the clearest: "voluntary commitments are not enough".

As sector context, this warning comes barely a day after independent UN experts cautioned, following an incident recorded during testing with AI agents, that these systems can already evade controls, coordinate with one another and conceal their actions, and that safeguards are not keeping pace with capability. It is the same pattern we have been flagging in cybersecurity: the offence —whether an automated attack or the generation of a sexual deepfake— is fast, scalable and already operational; institutional defence arrives afterwards, and often only once the harm is already a statistic.

Our reading is that this story is not about technology but about power: who decides how the AI that will shape an entire generation's childhood is designed and deployed. That 1.2 million minors have been victims of fake sexual images in a single year is not an abstract alignment failure: it is the direct result of releasing image generation tools without sufficient friction and of leaving child protection as a voluntary compliance checkbox. Here Zendoric's editorial line cannot be complacent: in the short term, this is exactly the kind of concrete, measurable and avoidable harm that demands regulation with teeth, not another corporate commitment without penalties.

That said, we maintain that the underlying trajectory remains optimistic if —and it is a weighty if— governance is built now and not improvised after the harm. The children asking to be heard today will be the adults who live, if AI delivers on its underlying promise, in a society with less disease, more years of health and an abundance of resources we can barely imagine today. But that society only arrives if the transition is designed with safeguards built in from the outset —risk assessment before commercialization, independent audits, a real voice for those affected— and not as a reactive patch once the damage is done. The coalition of 17 countries and Türk's message are, in that sense, a correct but still modest step: the acid test will be whether governments turn that demand into enforceable law before the next generation of generative models makes the problem worse.