Quick Answer
OpenAI has pledged $2 million for research on AI’s impact on children and signed a child safety charter at the Paris Peace Forum. The funding supports independent measurement work on AI’s risks and opportunities for children. Parents and schools should treat the pledge as a research commitment, not proof that existing AI tools are safe for unsupervised use.
Key Takeaways
- OpenAI announced a $2 million commitment on November 12, 2026.
- The funding supports research into AI’s risks and opportunities for children.
- OpenAI says participating research teams retain control over their conclusions.
- The wider Paris Peace Forum initiative includes more than 50 countries.
- ChatGPT says accounts identified as under 18 do not receive ads.
What did OpenAI pledge for child safety research?
OpenAI pledged $2 million to support scientific research on AI’s effects on children. The company announced the commitment on November 12, 2026, describing the recipient as the scientific research network under the Coalition for Children’s Rights and Protection in the Age of AI. OpenAI’s announcement says the funding will support shared measurement work intended to develop tools for studying both risks and opportunities.
The commitment matters because child safety questions need evidence that goes beyond a platform’s internal policies. Researchers need methods for examining how AI systems affect learning, emotional well-being, privacy, decision-making, and exposure to unsuitable material. A funding pledge does not establish what those effects are, but it can help create a more consistent way to study them.
OpenAI says research teams will retain control over their conclusions and that findings will be published even when results are negative or inconclusive. That condition is important because independent conclusions are more useful than research that only confirms a company’s preferred outcome. Parents, educators, and policymakers should watch for the published findings before treating the program as evidence that any specific chatbot is appropriate for a particular child.
What is the new AI child safety charter?
OpenAI says it signed the Charter of Engagement for Safer AI for Children during the Paris Peace Forum earlier in the week of November 12, 2026. The charter places OpenAI within a broader effort focused on children’s rights and protection as AI tools become more common in school, social, and consumer settings.
The Paris Peace Forum says the wider initiative involves more than 50 countries, seven international organizations, and 20 NGOs, philanthropies, and businesses. The same initiative includes funding commitments from OpenAI, Anthropic, Common Sense Media, and other AI-safety organizations for research and tools related to AI and child development. The Paris Peace Forum’s announcement puts the combined pledge at 8 million euros.
A charter and a funding commitment can establish shared priorities, but neither automatically creates enforceable rules for every AI product or country. The practical value depends on whether participating organizations publish measurable standards, disclose results, and make safety practices available to families and schools. Readers should distinguish voluntary commitments from laws, platform safeguards, and parental controls that are already available.
Why does research on children and AI matter?
Research on children and AI matters because children can use generative AI differently from adults, especially when a chatbot appears conversational, supportive, or authoritative. AI systems can produce useful explanations and creative help, but they can also provide inaccurate information, handle personal data, or generate responses that are unsuitable for a child’s maturity and circumstances.
Measurement work can help separate broad concerns from questions that researchers can test. For example, a study may examine whether an AI tool encourages disclosure of personal details, how it responds to distress, or whether a child understands that a generated answer can be wrong. The findings need to account for age, context, supervision, product design, and the type of task involved.
| Research area | Example question | Why it matters for children |
|---|---|---|
| Privacy | Does the tool encourage disclosure of personal information? | Children may not recognize the risks of sharing sensitive details. |
| Accuracy | Can a child identify when an AI-generated answer is wrong? | Confident errors can affect learning and decision-making. |
| Emotional well-being | How does the system respond when a child expresses distress? | Children need appropriate adult support for serious concerns. |
| Age appropriateness | Does the tool produce material suited to the child’s maturity? | Responses may be unsuitable depending on context and supervision. |
Parents should not assume that a general-purpose AI chatbot functions like a teacher, counselor, doctor, or child safety service. A child who uses an AI tool for schoolwork still needs guidance on checking sources, protecting personal information, and asking an adult for help with serious health, safety, or emotional concerns. Families considering social platform protections may also want to understand how AI age checks are being used to identify accounts that may belong to teens.
How does OpenAI describe ChatGPT protections for under-18 accounts?
OpenAI says ChatGPT does not show ads to accounts identified as belonging to people under 18. The company says it uses account-level age information and age-prediction signals to make that determination, according to its current advertising help documentation. OpenAI’s ChatGPT ads policy also says ads are not eligible to appear near personal health, mental health, or politics topics.
The under-18 ad policy addresses advertising exposure, not every safety issue a young person can encounter while using a chatbot. A child can still receive incomplete, misleading, or inappropriate generated information, depending on the prompt, the product settings, and the system’s behavior. Age estimation also depends on the information and signals available to a platform, so parents should not rely on it as a complete substitute for supervision.
The practical response is to discuss AI use directly with children and review the account settings available for the service they use. Families should also explain that private details, including full names, addresses, school schedules, passwords, medical information, and financial information, do not belong in prompts. That guidance reflects basic cybersecurity principles that apply to chatbots as well as email, games, and social platforms.
What does independent research mean in this program?
Independent research in OpenAI’s program means participating research teams retain control over their conclusions. OpenAI says the findings will be published, including results that are negative or inconclusive. That commitment gives the work more potential value because useful research must be able to identify limitations and harms rather than only documenting benefits.
Publication alone does not settle a complex safety question. Research quality depends on the methods, participant protections, sample design, data access, and whether other experts can review or replicate the findings. Child-focused research also requires particular care because children may not recognize privacy risks or be able to assess a chatbot’s reliability in the same way an adult can.
The most sensible interpretation is that OpenAI has funded a process for building evidence, not that it has delivered final answers. Parents and educators should look for clear public explanations of what researchers measure, what age groups are included, how findings are limited, and what actions the results support. Policy debates about AI developers can also evolve quickly, as shown by legal pressure on major AI companies over broader model safety concerns.
What should parents and schools do now?
Parents and schools should set clear boundaries for children’s AI use now, because research findings and voluntary commitments will take time to translate into product rules. The most useful boundaries cover supervision, privacy, fact-checking, and the types of topics that require an adult or qualified professional rather than a chatbot.
- Explain what AI can and cannot do. Tell children that a chatbot generates responses from patterns in data and can produce confident errors.
- Set a privacy rule. Keep names, addresses, passwords, account codes, health records, and private family information out of prompts.
- Review schoolwork together. Require children to check factual claims against reliable sources and follow their school’s rules on AI assistance.
- Use adult help for sensitive issues. Contact a parent, teacher, doctor, counselor, or emergency service when a child raises a health, safety, mental health, or abuse concern.
Schools should make their AI rules understandable enough for students and families to follow. A policy should state which tools are allowed, what information students must not submit, when disclosure is required, and how teachers will assess work. Schools should stop short of treating a chatbot as an authority on medical, mental health, legal, or emergency matters because generated responses cannot replace professional assessment.
What are the limits of OpenAI’s child safety commitment?
OpenAI’s child safety commitment has clear limits because the announcement describes funding and research support rather than a new universal product safeguard. The $2 million pledge can support studies and measurement tools, but it does not by itself prove that a specific AI system will respond safely in every child-related situation.
The wider initiative’s 8 million euro funding pledge also represents a coalition commitment, not a single global regulatory system. Countries, schools, platforms, and families use different rules and practices, and children encounter AI through many products beyond ChatGPT. The research may identify important patterns, but product teams and policymakers still need to decide how to apply the results.
For most families, the immediate priority remains informed supervision. Use age-appropriate settings where available, keep sensitive information out of prompts, and ask children to tell an adult when an AI response feels frightening, pressuring, or confusing. OpenAI’s pledge is a meaningful step toward better evidence, but safe use still depends on practical protections around the child.
FAQ
How much did OpenAI pledge for child AI safety research?
OpenAI pledged $2 million for the scientific research network under the Coalition for Children’s Rights and Protection in the Age of AI. OpenAI says the funding supports shared measurement work on AI’s risks and opportunities for children.
Did OpenAI sign a child safety agreement?
OpenAI says it signed the Charter of Engagement for Safer AI for Children during the Paris Peace Forum. The charter is part of a wider international child-wellbeing initiative involving governments, organizations, and businesses.
Does ChatGPT show ads to users under 18?
OpenAI says ChatGPT does not show ads to accounts identified as belonging to people under 18. The company says identification uses account-level age information and age-prediction signals, so parents should still supervise younger users.
Does the funding prove that AI is safe for children?
No, the funding does not prove that AI is safe for children. The pledge supports research intended to study risks and opportunities, while safe use still requires privacy protections, adult guidance, and appropriate platform rules.
What should a child do if an AI chatbot gives alarming advice?
A child should stop using the chatbot for that issue and tell a trusted adult. Parents should contact a qualified professional or emergency service when the response involves immediate danger, self-harm, abuse, medical symptoms, or another urgent safety concern.