France, Spain, Kenya and Australia Set 2026–27 Roadmap for Child-Safe AI

The Coalition's 2026–2027 Child-Safety Roadmap

At the high-level week of the 81st UN General Assembly, France, Spain, Kenya and Australia presented a 2026–2027 action plan for the Coalition for the Rights and Protection of Children in the Age of AI. The coalition was launched on 7 July 2026 by France, Spain and Kenya and later joined by Australia. It now brings together 23 states and eight UN bodies, including UNICEF, UNESCO, the International Telecommunication Union, the Office of the High Commissioner for Human Rights and the Committee on the Rights of the Child.

The most consequential idea in the roadmap is to move child-facing AI systems toward pre-market authorisation. The coalition states that AI systems claiming to deliver benefits to children should eventually be subject to prior approval before reaching the market. In the shorter term, it is asking for voluntary action: open access to AI models and data for independent researchers, concrete commitments from companies, and shared databases and evaluation methods to make independent safety checks possible.

Work will proceed along three axes. The first is keeping political ambition high through international milestones until 2027. The second is producing tangible results in thematic working groups covering safety and privacy by design, benchmarks and accountability, AI-generated child sexual abuse material and deepfakes, and AI education. The third is strengthening UN action through a common operational database of initiatives and evidence. The coalition also plans a global network of public, academic and private research centres and will ask companies, foundations and other partners to fund independent research on AI's impact on children. It expects to present its first voluntary commitments at the Paris Peace Forum on 10–11 November 2026.

How the Roadmap Pressures AI Developers and Regulators

Pre-market authorisation moves from idea to policy agenda

By stating that AI systems claiming benefits for children should eventually require prior authorisation before reaching the market, the coalition has turned a regulatory concept into a formal international work plan. This is not yet binding law: the immediate measures are voluntary while shared databases and evaluation methods are developed. But the direction is explicit, and the coalition is using the G7 Evian Call for a safer digital space for minors and Australia's platform duty-of-care framework as reference points.

France and Australia bring domestic bans into the global debate

The coalition's credibility rests partly on national measures already in force. President Macron cited France's ban on social media use for under-15s, while Prime Minister Albanese pointed to Australia's under-16 social media ban and the proposed My Feed, My Way bills to give users more control over algorithms. Those domestic rules are being presented as evidence that pre-emptive child protection is feasible and as templates for other coalition members.

AI developers face a transparency-for-trust trade-off

The interim demands are voluntary, but they are not cost-free. The coalition wants open access to models and data for independent researchers, plus funding for independent impact research. Companies that participate may gain legitimacy and early influence over child-safety benchmarks; those that stay out risk being associated with the AI-generated child sexual abuse material and deepfake risks the coalition has explicitly named. The practical question is whether major AI developers will open systems they have so far treated as proprietary.

The coalition must bridge very different digital realities

President Ruto's intervention sharpened a real tension. For many African children, he argued, access to technology means access to education, knowledge, healthcare and opportunity; the response cannot be to cut off access. That means the coalition's safety by design agenda has to accommodate jurisdictions where more access, not less, is the priority. This difference will shape how common benchmarks and pre-authorisation ideas are negotiated before the 2027 milestones.

What Governments and AI Developers Face Before the Paris Peace Forum

The roadmap gives different players a concrete calendar. These are the near-term decision points.

• AI developers with child-facing products: The coalition plans to present its first voluntary commitments at the Paris Peace Forum on 10–11 November 2026. Companies wanting a seat at that table have until then to decide whether to join the interim commitments on open model access and independent evaluation.
• Governments in the 23-state coalition: Align national AI safety rules with the three working groups—safety and privacy by design, benchmarks and accountability, and AI-generated child sexual abuse material and deepfakes—rather than duplicating parallel standards before the 2027 milestones.
• Independent researchers and civil society: The coalition's interim call for developer data and model access is the nearest lever for independent child-safety testing. Use it to request evaluation access to systems marketed as beneficial to children.
• Funders and foundations: The coalition is inviting financing for independent research on AI's impact on children and a global network of public, academic and private research centres. Funding announced before the Paris Forum will be visible in the first commitments cycle.