OpenAI’s $2 million for children’s AI safety research matters for one promise: publishing bad news

At the Paris Peace Forum, OpenAI signed the Charter of Engagement for Safer AI for Children and committed $2 million to the research network under the Coalition for Children’s Rights and Protection in the Age of AI. The money goes toward shared measurement: tools researchers can use to study AI’s risks and opportunities for children. The charter sets principles for age-appropriate AI plus a research program with independence safeguards. Research teams keep control of their conclusions, and findings will be published, including negative or inconclusive results.

OpenAI says this builds on its earlier call for global youth AI safety standards. It backs legislation with age-appropriate defaults, parental tools, regular risk assessments and accountability, and plans to engage European policymakers as the EU Kids Act develops. Earlier, it offered up to $5 million in grants for independent research on AI and teens aged 13 to 17 (our coverage).

Most public argument about kids and chatbots runs on anecdotes, lawsuits and worst-case headlines on one side, and glossy product promises on the other. What’s missing are shared, credible ways to measure what these systems do to young people over time. Funding that plumbing is less glamorous than a new safety feature and more durable. It complements OpenAI’s new violence and teen safeguards, which raise their own questions about age prediction and appeals (our analysis), and Common Sense Media’s finding that ChatGPT’s teen mode falls short (coverage).

A commitment to publish negative results is easy to write and hard to honor when the results concern your own product. If this network eventually produces findings that hurt and they’re published intact, the pledge will have meant something. The same independence question hangs over OpenAI’s approach to third-party safety assessments, where it wants scoped claims and room to remediate before publication (our analysis).

Two million dollars is small next to what frontier labs spend on compute or marketing. This is a contribution to a coalition, not a solution. OpenAI is also advocating for youth safety legislation that preserves access to its tools, a reasonable position that also serves its business. Both can be true.

The upside is real: AI tutoring, creative tools and help for kids who lack other support. So is the harm: dependency, inappropriate content, manipulation and developmental effects no one yet understands. If AI becomes a default companion for a generation before the evidence exists, the damage will be studied after the fact. Building the evidence now is the better order. More in our research coverage and OpenAI coverage.