Parents asking whether ChatGPT is safe for their children have had little independent evidence to go on. OpenAI's latest move is aimed at that gap: it has pledged $2 million to a research network studying how AI affects children's safety, wellbeing and development, and signed a charter on safer AI for children at the Paris Peace Forum. The money is small for OpenAI, but the question of who measures harm, and how, matters to every family that uses these tools.
The details come from coverage of the announcement, as OpenAI's own page could not be read. The funding goes to the research network of the Coalition for Children's Rights and Protection in the Age of AI, a French-led initiative, and is meant for shared measurement work: tools that let researchers assess risks and opportunities in comparable ways. Reports say OpenAI stated that independent researchers will keep control of their findings, including negative or inconclusive ones, and that research funding should come with stronger legal protections for children using AI products. It has not named grant recipients or a payment schedule.
The pledge sits inside a wider package. The Paris Peace Forum said AI developers including OpenAI and Anthropic, with child-safety groups, would commit €8 million, about $8.9 million, to research on AI and child development. Anthropic, Microsoft and Black Forest Labs also endorsed the charter. Separately, more than 50 countries backed a Children First Declaration covering health, education, protection and wellbeing in the digital age, with targets for 2045, and President Emmanuel Macron said no AI should reach children before being tested, evaluated and authorised.
The gap between the rhetoric and the mechanism is the story. Tech Times notes that no report names who would run the tests or who could refuse approval, that the Elysée describes the commitments as voluntary, and that the companies whose products would be tested are among the research funders. Independent evaluators have argued in a letter signed by more than 100 people that credible oversight needs testers the labs do not own or control.
What remains unknown is whether any of this will change the products. ChatGPT, Gemini and Claude are all used by teenagers, and none currently publishes child-specific test results a parent could look up. Until a standard and an approving body exist, a funded research network is a start on evidence, not a safety guarantee.