Cohere Responsible Use Policy

This document is a model card for Cohere's Command R and Command R+ language models, published as part of Cohere's responsible use documentation framework; it discloses model capabilities, safety evaluation results, intended use cases, and prohibited use cases, and operates in conjunction with Cohere's Usage Guidelines and customer agreements. The document states that Command R models are trained for text generation, summarization, code, markdown, RAG, and tool-use tasks across 23 languages, and that Cohere reserves the right to restrict API access at any time for violations of its Usage Guidelines and Terms of Use. The document explicitly states that Command R models are not recommended for decisions affecting access to financial services, employment, or housing without human oversight, and identifies astroturfing, misinformation generation, and human-outside-the-loop decision-making as prohibited use cases; it does not establish a formal enforcement procedure or appeals mechanism for API access restriction. The document engages with AI governance and ethics frameworks including bias evaluation using the BOLD dataset, and the prohibited use cases described intersect with emerging regulatory frameworks such as the EU AI Act, which addresses high-risk AI applications in employment, credit, and housing contexts. Compliance teams deploying Command R models in regulated sectors should evaluate whether their specific use cases fall within the prohibited categories identified in this document, particularly given the document's explicit warning against use in consequential individual decision-making.