OpenAI on Monday published a Child Safety Blueprint outlining how the company plans to integrate safeguards into AI systems for young users. The document, released by chief executive Sam Altman, describes a framework focused on age-appropriate design, content filtering, and collaboration with educators and parents.
The blueprint follows internal reviews and external feedback after concerns were raised about AI’s potential risks to minors. It commits OpenAI to restrict outputs that could harm children, including violent or sexually explicit material, while enabling beneficial uses such as homework assistance and language learning.
Altman stated the company will develop safety classifiers trained on child-specific data to detect harmful prompts before responses are generated. These systems will be tested in controlled environments before wider deployment.
The plan also calls for public transparency reports on incidents involving minors and annual reviews by an independent child safety panel. OpenAI will work with schools in the United States to pilot age-gated features and gather feedback from students aged 13 to 17.
Critics say the measures do not go far enough, noting that third-party developers can still build apps using OpenAI’s models without the same protections. The company acknowledges this gap and says it will explore technical and policy solutions to extend safeguards beyond its own products.
Source: openai.com