Our updated Preparedness Framework

OpenAI has released an updated version of its Preparedness Framework, which outlines how the company measures and mitigates severe risks from advanced AI models. According to the OpenAI Blog, the framework focuses on identifying capabilities that could lead to catastrophic harm, such as enabling cyberattacks, biological weapons, or autonomous AI systems that evade control. The update introduces clearer risk thresholds and stricter safety protocols for deploying models that approach these thresholds. This revision comes amid growing industry and regulatory scrutiny over frontier AI safety. For developers building AI workflows, the framework signals that future OpenAI models may come with more stringent access controls and evaluation requirements. While the framework itself is not a tool, it reinforces the importance of incorporating safety considerations into AI application design, especially for workflows that involve high-risk capabilities or sensitive domains.