OpenAI pledges $2 million to study how AI affects children

OpenAI has pledged $2 million to an international network studying how generative AI affects children. The company said on November 12 that it signed a child AI safety charter at the Paris Peace Forum and committed the money there. Its contribution will help researchers build shared tools for measuring the risks and benefits of AI use as children learn and develop.

Research teams will control their conclusions and publish findings even when the results are negative or inconclusive. The aim is to give families and schools evidence about which effects researchers have found, and in which age groups. OpenAI announced a funding commitment and a research programme. It did not announce changes to how children use ChatGPT.

A common way to measure the effects of AI

In its November 10 statement for the Child Summit, the Paris Peace Forum put the combined research and tool-development pledges at €8 million. It named AI developers including OpenAI and Anthropic and safety organizations including Common Sense Media. OpenAI’s $2 million disclosure two days later gives the amount of one company’s pledge. The charter lists Anthropic, Black Forest Labs, Microsoft and OpenAI as endorsers.

France, Spain and Kenya launched the Coalition for Children’s Rights and Protection in the Age of AI in July. Australia later joined as a co-chair. The coalition’s September 22 roadmap set out plans to connect public, academic and private research centres and seek funding and model access for independent studies. It identified the Paris forum as the venue for the first voluntary commitments.

The forum’s child AI initiative calls for research across age groups and developmental stages. A younger child may need a different explanation or safeguard from a teenager using the same tool. The research agenda covers concerns such as excessive use and emotional dependence, alongside opportunities in learning, creative work and accessibility. It includes studies that track development over time, as well as assessment of the risks and benefits of individual products.

A September report from UNICEF Innocenti illustrates why age matters. In consultations with 97 children aged 9 to 17 in five African countries, children described using chatbots for schoolwork and looking up information. Older adolescents reported using more advanced research tools. The report records their accounts of how they use AI. The proposed network would give researchers additional tools for studying the effects of that use.

What companies can fund and who controls the findings

The charter assigns industry funding to shared measurement. It reserves areas including developmental outcomes research and assessment standards for public or independent philanthropic funding, with separate accounts. Company researchers and technical staff can help develop the tools. Companies may raise documented objections to methods used to evaluate their own systems. Independent teams retain responsibility for the conclusions and decisions about publication.

The work will make methods available alongside results. That gives readers a basis for judging whether a finding from one service or age group applies elsewhere. The coalition plans to arrange access to models and relevant data for assessment. The charter calls for consent and data protection in those arrangements, and it gives sponsors no special access to children’s personal data.

France’s presidency also released a separate expert report on vulnerabilities involving generative AI on November 10. Nine experts produced 18 recommendations after 55 hearings with people from fields including health care, research, technology and child protection. The coalition’s September roadmap describes assessment before child-oriented systems reach the market as a longer-term aim. The charter sets out voluntary industry commitments that can help develop the evidence and methods for such decisions.

The research principles include hearing from children themselves. In a related effort, Everyone.AI, which works with the Paris Peace Forum on child AI issues, published a youth manifesto on June 15. It brought together 108 young contributors from 22 countries. They called for safeguards tailored to children’s development, clearer explanations from companies, and a role for young people in policy and product decisions.

The charter calls for the scientific committee’s chairs to publish an annual account of results, gaps in the evidence and changes to research priorities. It gives no date for the first report.