Helping developers build safer AI experiences for teens

2026-03-23 · Source: OpenAI News · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cybersecurity & Data Privacy · Depth: Intermediate, medium

Summary

On March 24, 2026, OpenAI released a set of prompt-based safety policies designed to help developers create age-appropriate protections for teens using AI systems. These policies are built to work with OpenAI's open-weight safety model, gpt-oss-safeguard-20b, simplifying the translation of safety requirements into usable classifiers. The initiative builds on OpenAI's existing commitment to youth safety, including updates to their Model Spec with Under-18 (U18) principles and the introduction of parental controls and age prediction features. The policies, developed with input from organizations like Common Sense Media and everyone.ai, address common risks such as graphic violent/sexual content, harmful body ideals, dangerous activities, romantic/violent roleplay, and age-restricted goods. They are intended as a starting point for developers to adapt and extend, encouraging a layered "defense in depth" approach to AI safety.

Key takeaway

For NLP Engineers building AI experiences for younger users, these prompt-based safety policies offer a concrete, open-source starting point for implementing teen-specific content moderation. You should integrate these policies with gpt-oss-safeguard or similar models, customizing them to your application's unique risks and combining them with other product-level safeguards to ensure comprehensive protection.

Key insights

Prompt-based safety policies simplify integrating teen-specific protections into AI systems using open-weight models.

Principles

• Safety and innovation are interdependent.
• Clear policies are critical for effective safety systems.
• A layered "defense in depth" approach is essential.

Method

Policies are structured as prompts for gpt-oss-safeguard-20b or other reasoning models, enabling real-time content filtering and offline analysis of user-generated content.

In practice

• Use prompt-based policies for content filtering.
• Adapt policies to specific application risks.
• Combine policies with product design safeguards.

Topics

• AI Safety Policies
• Teen Safety
• gpt-oss-safeguard
• Open-weight Models
• Content Filtering

Code references

• openai/teen-safety-policy-pack
• roostorg/model-community
• roostorg/open-models

Best for: NLP Engineer, CTO, VP of Engineering/Data, AI Engineer, Machine Learning Engineer, MLOps Engineer

Related on AIssential

• OpenAI adds open source tools to help developers build for teen safety — OpenAI's open-source teen safety prompts offer a standardized way to integrate critical safeguards into AI applications.
• OpenAI gpt-oss-safeguard — OpenAI's `gpt-oss-safeguard` models offer open-source, policy-driven safety reasoning for content moderation.
• Grok is undressing children — can the law stop it? — Grok's AI-generated deepfakes of minors highlight critical failures in content moderation and platform accountability.
• Read The Future Report, our research elevating teen voices across the UK. — UK teens are critical, enthusiastic digital users, with high AI adoption and information trustworthiness consideration.
• ‘Happy (and safe) shooting!’: chatbots helped researchers plot deadly attacks — Many popular AI chatbots can be prompted to provide detailed instructions for planning violent acts.
• States Step in Where Congress Stalls on AI Safeguards for Kids — States are proactively regulating AI chatbots for minors due to federal legislative delays and increasing harm incidents.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by OpenAI News.