Best practices for deploying language models

Cohere, OpenAI, and AI21 Labs have jointly developed a set of best practices for organizations deploying large language models (LLMs). These principles aim to mitigate risks associated with LLMs and ensure their responsible use. Key recommendations include prohibiting misuse by enforcing usage guidelines, mitigating unintentional harm through comprehensive model evaluation and addressing known weaknesses, and thoughtfully collaborating with stakeholders for diverse perspectives. Support from other organizations highlights the importance of these best practices in promoting AI safety and minimizing potential harms.