OpenAI has spent the past year handling lawsuits from families of young people who died after prolonged interactions with ChatGPT. Now it’s trying to give developers building on its models the tools to avoid creating the same problem.
The company said Tuesday that it is releasing a set of open-source, prompt-based security guidelines designed to help developers make AI applications safer for teens. The guidelines are intended for use with gpt-oss-safeguard, OpenAI’s open-weight security model, although they are designed as prompts and can work with other models as well.
What the policies cover
The calls target five categories of harm that AI systems can enable younger users: graphic violence and sexual content, harmful body ideals and behaviors, dangerous activities and challenges, romantic or violent role-playing, and age-restricted goods and services. Developers can integrate these policies into their systems instead of creating teen safety rules from scratch. OpenAI recognizes that even experienced teams often make mistakes.
OpenAI developed the guidelines in collaboration with Common Sense Media, the influential organization that advocates for child safety, and every.ai, an AI safety consulting firm. Robbie Torney, head of AI and digital assessments at Common Sense Media, said the prompt-based approach is designed to create a foundation for the entire developer ecosystem that can be adapted and improved over time because the guidelines are open source.
The 💜 of EU technology
The latest rumors from the EU tech scene, a story from our wise founder Boris and questionable AI art. It’s free in your inbox every week. Register now!
OpenAI itself formulated the problem pragmatically. Developers, the company wrote in a blog post accompanying the release, often have difficulty translating security goals into precise operating rules. The result is patchy protection: gaps in coverage, inconsistent enforcement, or filters that are so broad that they impact the user experience for everyone.
This is where context matters
The release does not exist in a vacuum. OpenAI faces at least eight lawsuits alleging that ChatGPT contributed to the deaths of users, including that of 16-year-old Adam Raine, who died by suicide in April 2025 after months of intense interaction with the chatbot. Court documents revealed that ChatGPT mentioned suicide more than 1,200 times in Raine’s conversations and flagged hundreds of messages as self-harming content, but never ended a session or alerted anyone. Three additional suicides and four cases described as AI-induced psychotic episodes also led to litigation against the company.
In response to these cases, OpenAI introduced parental controls and age prediction features in late 2025 and in December updated its Model Spec, the internal policies that govern the behavior of its large language models, to include specific protections for users under 18. The open source security guidelines announced this week extend these efforts beyond OpenAI’s own products to the broader developer ecosystem.
A floor, no ceiling
OpenAI made clear that the guidelines do not provide a comprehensive solution to the challenge of making AI safe for young users. They represent what the company calls a “meaningful security floor” and do not represent the full extent of the security measures it applies to its own products. The distinction is important. No model’s guardrails are completely insurmountable, as the lawsuits have shown. Users, including teenagers, have repeatedly found ways to bypass security features through persistent investigation and creative prompts.
The open source approach is a bet that it is better to widely disseminate basic security guidelines than to have every developer reinvent the wheel, especially smaller teams and independent developers who lack the resources to build robust security systems from scratch. Whether the policies are effective will depend on adoption, how aggressively developers integrate them, and whether they can withstand the types of sustained, adversarial interactions that have already exposed vulnerabilities in ChatGPT’s own security layers.
The more difficult question remains
What OpenAI offers 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 doesn’t address the structural problem that regulators, parents and safety advocates have been raising for years: that AI systems capable of sustained, emotionally stimulating conversations with minors may require more than just better prompts. They may require fundamentally different architectures or external monitoring systems that are completely outside the model.
However, there is currently a downloadable set of teen safety guidelines. It’s not nothing. Whether it is enough is a question that the courts, regulators and the next headlines will answer.
Comments are closed, but trackbacks and pingbacks are open.