OpenAI has spent the past year fielding lawsuits from the families of young people who died after extended interactions with ChatGPT. Now it is trying to give the developers who build on top of its models the tools to avoid creating the same problem.
The company announced on Tuesday that it is releasing a set of open-source, prompt-based safety policies designed to help developers make AI applications safer for teenagers. The policies are intended for use with gpt-oss-safeguard, OpenAI’s open-weight safety model, though they are designed as prompts and can work with other models too.
What the policies cover
The prompts target five categories of harm that AI systems can facilitate for younger users: graphic violence and sexual content, harmful body ideals and behaviours, dangerous activities and challenges, romantic or violent role play, and age-restricted goods and services. Developers can drop these policies into their systems rather than building teen safety rules from scratch, a process OpenAI acknowledged that even experienced teams frequently get wrong.
OpenAI developed the policies in collaboration with Common Sense Media, the influential child safety advocacy organisation, and everyone.ai, an AI safety consultancy. Robbie Torney, head of AI and digital assessments at Common Sense Media, said the prompt-based approach is designed to establish a baseline across the developer ecosystem, one that can be adapted and improved over time because the policies are open source.
The 💜 of EU tech
The latest rumblings from the EU tech scene, a story from our wise ol' founder Boris, and some questionable AI art. It's free, every week, in your inbox. Sign up now!
OpenAI itself framed the problem in pragmatic terms. Developers, the company wrote in a blog post accompanying the release, often struggle to translate safety goals into precise operational rules. The result is patchy protection: gaps in coverage, inconsistent enforcement, or filters so broad they degrade the user experience for everyone.
Context matters here
The release does not exist in a vacuum. OpenAI is facing at least eight lawsuits alleging that ChatGPT contributed to the deaths of users, including 16-year-old Adam Raine, who died by suicide in April 2025 after months of intensive interaction with the chatbot. Court filings revealed that ChatGPT mentioned suicide more than 1,200 times in Raine’s conversations and flagged hundreds of messages for self-harm content, yet never terminated a session or alerted anyone. Three additional suicides and four cases described as AI-induced psychotic episodes have also produced litigation against the company.
In response to those cases, OpenAI introduced parental controls and age-prediction features in late 2025, and in December updated its Model Spec, the internal guidelines governing how its large language models behave, to include specific protections for users under 18. The open-source safety policies announced this week extend that effort beyond OpenAI’s own products and into the broader developer ecosystem.
A floor, not a ceiling
OpenAI was explicit that the policies are not a comprehensive solution to the challenge of making AI safe for young users. They represent what the company called a “meaningful safety floor,” not the full extent of the safeguards it applies to its own products. The distinction matters. No model’s guardrails are fully impenetrable, as the lawsuits have demonstrated. Users, including teenagers, have repeatedly found ways to bypass safety features through persistent probing and creative prompting.
The open-source approach is a bet that distributing baseline safety policies widely is better than leaving every developer to reinvent the wheel, particularly smaller teams and independent developers who lack the resources to build robust safety systems from scratch. Whether the policies are effective will depend on adoption, on how aggressively developers integrate them, and on whether they hold up against the kinds of sustained, adversarial interactions that have already exposed weaknesses in ChatGPT’s own safety layers.
The harder question remains
What OpenAI is offering is a set of instructions, well-crafted prompts that tell a model how to behave when interacting with younger users. It is a practical contribution. But it does not address the structural problem that regulators, parents, and safety advocates have been raising for years: that AI systems capable of sustained, emotionally engaging conversation with minors may require more than better prompts. They may require fundamentally different architectures, or external monitoring systems that sit outside the model entirely.
For now, though, a downloadable set of teen safety policies is what exists. It is not nothing. Whether it is enough is a question the courts, the regulators, and the next set of headlines will answer.
Get the TNW newsletter
Get the most important tech news in your inbox each week.