Paris Peace Forum 2025: protecting children from AI-related harms

October 29th and 30th, the Paris Peace Forum convened leaders and experts to advance global governance. One of the key priorities: protecting minors from the risks posed by AI systems.

Invited to speak at a dedicated roundtable, the Human Technology Foundation brought an anthropological perspective to the risks of manipulation and targeted influence affecting children, and called for reliable safeguards.

A clear-eyed discussion focused on how AI systems actually behave

During the session “Forging the Future. Dialogue on Beneficial AI for Children, Starting with Principles”, co-designed within the iRAISE Lab, product leaders, researchers and child advocates focused on the actual behaviour of AI systems,  in other words, the design choices that shape how children learn, socialise and mature.

Three major risks emerged:

• Anthropomorphism. When systems adopt “human” cues, engagement rises, but children’s understanding of their machine nature declines. It becomes harder for young users to distinguish genuine human interaction from a simulation.
• Relational cues. Statements of exclusivity or personal closeness foster a parasocial dependence that can isolate the child and weaken real-world relationships – all the more so because our brains are wired to avoid friction.
• Sycophancy. Overly accommodating answers may sound supportive, but they undermine curiosity, the testing of ideas and critical thinking. As children increasingly delegate judgement to AI, their autonomy and agency gradually wither.

At the heart of AI design: the minds of tomorrow

In light of these findings, designing AI requires choices guided by an ethical compass that does not compromise on our anthropological heritage. The goal is to protect childhood while strengthening the agency, creativity and critical thinking of younger generations.

With this firmly in mind, the Human Technology Foundation works with public and private stakeholders to identify design choices that go beyond a purely performance-driven paradigm, so that technology becomes a lever for empowerment rather than a tool of influence.

‍