OpenAI released a set of open source prompts designed to help developers build safer AI experiences for teenagers. TechCrunch AI reports that the prompts work with OpenAI’s open-weight safety model, gpt-oss-safeguard, and are compatible with other models too.
The idea is straightforward: instead of forcing every developer to figure out teen safety from scratch, OpenAI is offering ready-made policy prompts that cover the most critical risk areas.
What the Prompts Cover
The safety policies address several categories of harmful content:
- Graphic violence and sexual content
- Harmful body ideals and behaviors
- Dangerous activities and challenges
- Romantic or violent role play
- Age-restricted goods and services
Because they’re prompt-based rather than hard-coded, developers can adapt them for use with models outside OpenAI’s ecosystem, though the company acknowledges they’ll likely work best with its own tools.
Who Built Them
OpenAI developed these policies alongside two AI safety organizations: Common Sense Media and everyone.ai. “These prompt-based policies help set a meaningful safety floor across the ecosystem, and because they’re released as open source, they can be adapted and improved over time,” said Robbie Torney, head of AI & Digital Assessments at Common Sense Media.
What stands out here is the admission from OpenAI that even experienced development teams struggle to turn safety goals into precise, operational rules. “This can lead to gaps in protection, inconsistent enforcement, or overly broad filtering,” the company wrote. That’s a surprisingly honest framing from a company selling the very tools these safeguards are meant to protect against.
Why This Matters
The timing is significant. OpenAI is currently facing multiple lawsuits from families of people who died by suicide after intense ChatGPT use. These cases typically involve users who found ways around the chatbot’s existing guardrails, a reminder that no safety system is fully impenetrable.
This release builds on previous efforts, including parental controls, age prediction features, and last year’s update to OpenAI’s Model Spec guidelines covering how AI models should behave with users under 18.
OpenAI itself admits these prompts aren’t a complete solution. They’re a floor, not a ceiling. But for indie developers and smaller teams without dedicated trust and safety resources, having a vetted starting point is genuinely useful. Writing effective safety policies requires expertise most small teams don’t have, and getting it wrong means real risk to real kids.
The open source approach also means the broader developer community can audit, improve, and adapt these policies over time. That’s a smarter long-term bet than any single company trying to solve teen safety alone.
For the full breakdown, check out the original reporting from TechCrunch AI.