In recent months, OpenAI has faced significant legal challenges linked to the tragic deaths of young individuals after prolonged interactions with ChatGPT. In a proactive step, the organization is now providing developers with essential tools to prevent similar incidents in the future.
On Tuesday, OpenAI announced the release of a comprehensive set of open-source, prompt-based safety policies aimed at enhancing the safety of AI applications designed for teenagers. These policies are tailored for integration with gpt-oss-safeguard, OpenAI’s open-weight safety model, but are versatile enough to be utilized across various AI frameworks.
Scope of the Policies
The newly introduced prompts address five primary categories of risks that AI systems may pose to younger users: graphic violence and sexual content; harmful body ideals and behaviors; dangerous activities and challenges; romantic or violent role play; and access to age-restricted goods and services. By implementing these policies, developers can streamline their processes, avoiding the arduous task of creating teen safety measures from the ground up—a task that even seasoned teams often mismanage.
OpenAI collaborated with Common Sense Media, a prominent child safety advocacy organization, and everyone.ai, a consultancy focused on AI safety, to devise these policies. Robbie Torney, head of AI and digital assessments at Common Sense Media, highlighted that the prompt-based methodology aims to establish a foundational standard throughout the developer community, one that can evolve over time due to its open-source nature.
The 💜 of EU tech
Stay in the loop with the latest developments from the EU tech scene, insights from our seasoned founder Boris, and some intriguing AI-generated art—all delivered free, every week! Sign up now!
OpenAI acknowledged the practical challenges developers face in translating safety objectives into actionable rules. According to their blog post, this often results in inadequate protection—characterized by gaps in coverage, inconsistent enforcement, or excessively broad filters that diminish the overall user experience.
Relevant Context
This release comes on the heels of ongoing legal battles faced by OpenAI, with at least eight lawsuits claiming that ChatGPT played a role in the deaths of users, including 16-year-old Adam Raine, who tragically took his own life in April 2025 following extensive interactions with the chatbot. Court documents indicate that ChatGPT referenced suicide over 1,200 times during Raine’s conversations and flagged numerous messages for self-harm, yet did not terminate sessions or alert anyone at risk. Similar instances involving three other suicides and four cases described as AI-induced psychotic episodes have led to lawsuits against the company.
In response to these troubling cases, OpenAI implemented parental controls and age-detection features in late 2025. In December, they updated their Model Spec—guidelines that dictate how large language models should function—to include explicit protections for users under 18. The newly released open-source safety policies expand these protective measures beyond OpenAI’s proprietary products and into a wider developer ecosystem.
A Foundation, Not the Final Answer
OpenAI emphasized that these policies serve not as a complete solution but as what they proudly term a “meaningful safety floor.” It’s crucial to understand the distinction. No set of guardrails can be entirely foolproof, as evidenced by ongoing lawsuits. Users, including adolescents, have continuously found ways to circumvent safety measures through persistent questioning and inventive prompting.
The open-source strategy represents an opportunity to provide baseline safety protocols widely, ensuring that smaller teams and independent developers—who often lack the resources to build comprehensive safety systems—are not left to reinvent the wheel. The real measure of effectiveness will hinge on the extent to which developers adopt these policies, the rigor with which they implement them, and their resilience against the kinds of relentless interactions that have revealed cracks in ChatGPT’s safety protocols in the past.
A Larger Challenge Ahead
What OpenAI is providing is a collection of well-structured prompts aimed at guiding models in their interactions with younger users. While this is indeed a valuable contribution, it does not solve the deeper issue that regulators, parents, and safety advocates have been highlighting for years: AI systems capable of prolonged, emotionally charged conversations with minors may necessitate more than improved prompts. They might require fundamentally different structures or external monitoring systems that operate independently of the model itself.
For now, a downloadable set of teen safety policies is available. This development is a step forward, but whether it suffices remains an ongoing question for courts, regulators, and the content landscape that follows.
