The tech world is witnessing rapid development in the field of artificial intelligence, which imposes new challenges regarding user safety, especially among younger age groups. In this context, OpenAI announced on Monday the launch of a new set of open-source safety prompts designed specifically to help developers build AI applications with built-in protection mechanisms tailored for teenage users. This step comes at a time of growing concerns about the impact of modern technologies on the mental and physical health of youth.
Strategic collaboration to address digital risks
These prompts and policies were developed in close collaboration with the non-profit organization Common Sense Media, as well as the Every1.ai foundation. This package provides practical and actionable guidance for dealing with common risks teenagers face in the digital environment. These risks include exposure to sexually explicit content, graphic violence, the promotion of harmful body image standards that may affect teenagers’ self-confidence, as well as incitement to dangerous activities and the promotion of age-restricted goods.
Advanced technology for content processing
To ensure the effectiveness of these prompts, the new package was designed to work in full compatibility with OpenAI’s “gpt-oss-safeguards” model. Released in October 2025, this model is an open-weights inference system that allows developers to apply custom safety policies during the inference and data processing stage, rather than relying entirely on traditional pre-trained classifiers that often lack sufficient flexibility to handle complex contexts.
Unlike traditional content moderation tools that rely on rigid blocklists, this model features the ability to directly interpret the policies defined by the developer. Not only that, but the system provides transparent and clear inferences and justifications for the decisions it makes, giving developers a deeper understanding of why specific content was blocked or allowed.
Challenges of translating safety goals into precise rules
OpenAI acknowledged that translating broad safety goals into precise and enforceable rules has been and continues to be an ongoing challenge in the tech industry. In an official statement, the company explained: “Even seasoned teams often find it extremely difficult to translate comprehensive safety goals into precise and actionable rules, especially since this requires deep subject matter expertise and a thorough understanding of AI mechanics.”
Meanwhile, Robbie Torney, who leads AI and digital ratings at Common Sense Media, noted that the new policies can establish meaningful safety ground across the entire AI ecosystem. He emphasized that this step addresses a major operational gap that left many third-party developers for years without clear guidelines on how to protect young users. Developers can now access this advanced model via the Hugging Face platform, and they can find the new prompt package on GitHub.
A series of initiatives to protect youth
This release is the latest in a series of successive moves taken by OpenAI to address growing concerns about youth safety. In September 2025, the company introduced ChatGPT parental control tools, followed in December 2025 by updating its model specifications to include specific guidelines for users under 18 years of age. Earlier this year, the company began deploying age-prediction technology to automatically apply teen protection safeguards.
Alongside technical solutions, OpenAI has partnered strategically with Common Sense Media to support a legislative initiative in California known as the “Kid Safe AI Act,” which aims to establish enforceable legal protections for teenagers and young people when using artificial intelligence systems.
Legal pressures and future aspirations
These intensified efforts come amid mounting legal and regulatory pressures on tech companies. OpenAI is currently facing multiple lawsuits alleging that ChatGPT has contributed to harming young users, most notably a wrongful death lawsuit filed against the company by the parents of a teenager in California.
Despite all these efforts and developments, OpenAI issued a clear warning that the new safety prompt package does not represent a final solution or an absolute guarantee of teen safety. Instead, the company described this step merely as a baseline and starting point that developers can customize and improve over time as technology evolves and new challenges emerge, emphasizing that building a safe digital environment requires ongoing work and regular updates to the policies and tools used.
Frequently Asked Questions
What are the safety prompts launched by OpenAI?
They are a set of open-source guidelines and policies designed to help developers build artificial intelligence applications that include built-in protections for youth and teenagers.
What are the most prominent risks these prompts aim to mitigate?
The prompts aim to reduce sexually explicit content, graphic violence, harmful body image standards, dangerous activities, and the promotion of age-restricted goods.
How does the gpt-oss-safeguards model work?
It operates as an open-weights inference system that allows developers to apply custom safety policies during inference, providing transparent explanations for its decisions instead of relying on pre-trained classifiers.
Are these prompts considered a final solution to ensure teen safety?
No, OpenAI has clarified that these prompts represent a baseline and starting point that developers can customize and improve over time as technology evolves.
