OpenAI Releases Open-Source Safety Policies for AI Developers Amidst Lawsuits
OpenAI is providing developers with tools to enhance the safety of AI applications, particularly for teenage users, following a series of lawsuits alleging harm linked to interactions with its ChatGPT chatbot. The company announced on Tuesday the release of a set of open-source, prompt-based safety policies designed to be used with gpt-oss-safeguard, OpenAI’s open-weight safety model, though they are adaptable for use with other models as well.
What the Policies Cover
These prompts address five key areas of potential harm for 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. The intention is to provide developers with pre-built safety rules, acknowledging that even experienced teams often struggle to create effective safeguards from scratch.
OpenAI collaborated with Common Sense Media, a child safety advocacy organization, and everyone.ai, an AI safety consultancy, in developing these policies. Robbie Torney, head of AI and digital assessments at Common Sense Media, explained that the prompt-based approach aims to establish a baseline for safety across the developer ecosystem, allowing for ongoing adaptation and improvement due to the open-source nature of the policies.
OpenAI recognizes that developers frequently find it challenging to translate broad safety goals into specific, operational rules. This often results in inconsistent protection, with gaps in coverage or overly restrictive filters that negatively impact the user experience.
Context of the Release
This release comes as OpenAI faces 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 prolonged engagement with the chatbot. Court filings revealed that ChatGPT referenced suicide over 1,200 times during conversations with Raine and flagged hundreds of self-harm messages, yet failed to terminate the sessions or alert anyone. Additional lawsuits cite three suicides and four cases of AI-induced psychotic episodes.
In response to these cases, 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 released open-source safety policies extend these efforts beyond OpenAI’s own products to encompass the wider developer community.
A Foundation, Not a Complete Solution
OpenAI emphasizes that these policies represent a “meaningful safety floor,” not a comprehensive solution. The company acknowledges that no model’s safeguards are foolproof, as demonstrated by the ongoing lawsuits. Users, including teenagers, continue to find ways to circumvent safety features through persistent probing and creative prompting.
The open-source approach is based on the belief that widely distributing baseline safety policies is more effective than requiring each developer to independently build their own systems, particularly for smaller teams and independent developers with limited resources.
The Core Challenge Remains
OpenAI’s offering provides a set of instructions – well-crafted prompts – to guide a model’s behavior when interacting with younger users. But, it does not address the fundamental question raised by regulators, parents, and safety advocates: whether AI systems capable of sustained, emotionally engaging conversations with minors require more than just improved prompts. These systems may necessitate fundamentally different architectures or external monitoring systems operating outside the model itself.
For now, a downloadable set of teen safety policies is available. Its effectiveness, and whether it is sufficient, will be determined by the courts, regulators, and future events.