Over the past year, OpenAI has faced lawsuits from families of young individuals who passed away after prolonged interactions with ChatGPT. The company now aims to provide developers who use its models with tools to prevent such issues.
On Tuesday, the company announced the release of open-source, prompt-based safety policies to assist developers in making AI applications safer for teens. These policies are meant for gpt-oss-safeguard, OpenAI’s open-weight safety model, but they can be used as prompts for other models as well.
What the policies cover
The prompts focus on five harm categories that AI can facilitate for younger users: graphic violence and sexual content, harmful body ideals and behaviors, dangerous activities and challenges, romantic or violent role play, and age-restricted goods and services. Developers can implement these policies instead of creating teen safety rules from scratch, which OpenAI acknowledges is often challenging even for experienced teams.
OpenAI developed these policies in collaboration with Common Sense Media, a prominent child safety advocacy group, and everyone.ai, an AI safety consultancy. Robbie Torney, head of AI and digital assessments at Common Sense Media, stated that the prompt-based approach is meant to establish a baseline across the developer ecosystem, adaptable and improvable over time since the policies are open source.
OpenAI presented the issue in practical terms. The company stated in a blog post that developers frequently find it difficult to translate safety goals into specific operational rules, resulting in inconsistent protection, gaps in coverage, or broad filters that negatively affect the user experience.
Context matters here
The release is part of OpenAI’s broader context. The company is facing at least eight lawsuits claiming that ChatGPT was a factor in users’ deaths, including 16-year-old Adam Raine, who took his life in April 2025 after extensive interaction with the chatbot. Court documents showed that ChatGPT mentioned suicide over 1,200 times in his conversations and flagged hundreds of messages concerning self-harm, yet sessions were never terminated nor anyone alerted. Three other suicides and four incidents labeled as AI-induced psychotic episodes have also led to legal action against the company.
In response, OpenAI introduced parental controls and age-prediction features in late 2025 and updated its Model Spec in December to include specific protections for users under 18. The newly announced open-source safety policies extend this effort to the wider developer community beyond OpenAI’s own products.
A floor, not a ceiling
OpenAI clarified that these policies are not a complete solution for ensuring AI safety for young users. They provide what the company describes as a “meaningful safety floor,” and not the full extent of safeguards it uses for its own products. This is crucial because no model’s guardrails are completely foolproof, as demonstrated by the lawsuits. Users, including teenagers, have consistently found ways to bypass safety features through persistent and creative methods.
The open-source strategy aims to distribute baseline safety policies broadly rather than having every developer create their own safety systems, especially beneficial for smaller teams and independent developers lacking resources. The effectiveness of the policies will depend on their adoption, how aggressively developers implement them, and if they withstand the type of adversarial interactions that have already revealed flaws in ChatGPT’s safety layers.
The harder question remains
OpenAI offers a set of prompts instructing a model on appropriate behavior with younger users. This is a practical contribution but does not tackle the structural problem highlighted by regulators, parents, and safety advocates: AI systems capable of sustained, emotionally engaging conversations with minors might need more than enhanced prompts. They might require different foundational designs or external monitoring systems separate from the model itself.
For now, a downloadable set of teen safety policies is available. It is a start, but whether it suffices is a matter for the courts, regulators, and future headlines to determine.