A coalition led by children's rights group 5Rights Foundation wants governments to enforce pre market child safety duties under existing UN commitments, not new treaty law.
A coalition of more than 100 organizations says AI companies should have to prove their products are safe for children before launch, not after harm has already occurred. The call, released Monday by 5Rights Foundation and its co-signatories including Amnesty International and Save the Children, lands one day before the UN's inaugural Global Dialogue on AI Governance.
The coalition is not asking for new international AI law. The request is for governments to apply obligations they have already accepted under the UN Convention on the Rights of the Child and the Global Digital Compact to AI products aimed at minors. The practical effect is a burden-of-proof shift: companies demonstrate safety first; children, parents, and regulators stop being the testbed.
The UN Convention on the Rights of the Child is the most widely ratified human-rights treaty in history. Its obligations include protections from economic exploitation, from harmful information, and from arbitrary interference with privacy. The coalition's argument is that these obligations already cover what AI products aimed at children are doing. The enforcement is missing.
The 5Rights-led coalition statement names ten concrete levers it wants governments to adopt. They include pre-market safety requirements for AI products marketed to minors, financial penalties for child-rights violations, bans on design features engineered to exploit young users' psychological vulnerabilities, restrictions on commercial use of children's biometric data, and a duty on companies to assess and disclose child-safety risks during model development. The ask is a defensive perimeter, not a guideline.
Engagement and data extraction are what AI products aimed at children are rewarded for; child safety is treated as a cost. 5Rights executive director Leanda Barrington-Leach, on the record with the coalition statement, said children themselves have already diagnosed the problem. Respecting children's rights, the coalition argues, has to become a condition of doing business, not an optional addition.
Lawsuits have been filed against Character Technologies and OpenAI over companion chatbots marketed to children as emotionally relational products. The cases are pending, not adjudicated. They give the coalition's argument concrete weight: products designed to draw children in, harms that were foreseeable, and a regulatory framework that has so far arrived after the damage.
Routing child-safety enforcement through frameworks that already exist is an unusual choice. It lowers the political cost of compliance and answers the standard objection that AI treaties take too long to negotiate. Treaties are not the point. The point is whether governments do what they have already promised.
The dialogue's closing statement will test that bet. If the language stays at the level of aspiration, the coalition's case moves to national regulators and courts, where the convention is already enforceable. If the language sharpens to mandatory pre-market obligations, the burden-of-proof shift the coalition is asking for starts to land.