Cohere Usage Policy

This document is Cohere's Usage Policy (also referred to as an Acceptable Use Policy), which governs permitted and prohibited uses of Cohere's API, models, and related services, with its legal basis deriving from Cohere's broader Terms of Service. The policy states that developers must outline and obtain approval for their use case to access the Cohere API, must document potential harms of their applications, and must refer to model cards for information about model capabilities and limitations. The policy enumerates categories of strictly prohibited use including generation of content that facilitates violence or physical harm, CSAM, hate speech, harassment, fraud, privacy violations, disinformation, influence operations, unauthorized surveillance, and applications targeting minors without appropriate safeguards; it also identifies a distinct set of use cases requiring special approval, such as medical or health-related applications, legal advice, applications targeting or used by minors, and high-stakes automated decision-making affecting individuals. The policy engages with the EU AI Act, which classifies certain AI applications as high-risk or prohibited, and with applicable consumer protection and data privacy frameworks including GDPR and CCPA, particularly where developers deploy Cohere models in health, legal, financial, and minor-facing contexts. Compliance teams evaluating downstream deployment of Cohere's API should note that the policy places affirmative obligations on developers as operators, including harm documentation, use case approval, and safeguard implementation, creating a tiered accountability structure between Cohere and its API customers.