This system card analyzes the latest large language model in the GPT family of models, GPT-4, and the potential societal impacts of LLMs in our daily lives. The safety challenges presented by the models limitations and capabilities are highlighted, along with the safety processes adopted by OpenAI to prepare GPT-4 for deployment. While these mitigations and processes alter GPT-4s behavior and prevent certain kinds of misuses, they have limitations and point to the need for anticipatory planning and governance. The system card also discusses the training process of GPT models and the interventions implemented to mitigate potential harms from GPT-4s deployment. Content warning: This document contains sexual, hateful, or violent content that some may find disturbing or offensive.
