Third-party AI models
Third-party AI models
Coveo uses third-party AI models, including large language models (LLMs), to power AI services within the Coveo Platform, such as Relevance Generative Answering (RGA). Your use of the AI services must comply with the applicable third-party terms listed below, including any usage restrictions and supported regions.
Microsoft Azure OpenAI Service (OpenAI model families)
Coveo may use OpenAI model families provided through Microsoft Azure OpenAI Service. Applicable terms include Microsoft Acceptable Use Policy.
Amazon Bedrock (AWS)
Coveo may use Amazon Bedrock to access foundation model families, including:
-
AWS Titan. Applicable terms include AWS Responsible AI Policy and AWS Acceptable Use Policy.
-
AWS Nova. Applicable terms include AWS Responsible AI Policy and AWS Acceptable Use Policy.
-
Anthropic Claude. Applicable terms include Anthropic’s Usage Policy and its supported countries and regions.
-
Meta Llama. Applicable policies include Meta’s use policies for LLama model families:
|
Customer data handling with AWS Bedrock
When Coveo uses AWS Bedrock to access third-party AI models, all inference operations run on AWS-managed instances of those models within AWS’s secure infrastructure. Customer data is processed within AWS and isn’t shared with third-party AI model providers. Customer data used through AWS Bedrock isn’t used to train or improve any original third-party foundation or base models. |
Hugging Face models
Coveo may use select open-source model families available through Hugging Face to power or enhance certain Coveo Platform capabilities. These models are hosted locally on AWS’s secure environment, and no customer data is shared with third-party AI model providers.
Restrictions and updates
The list of third-party AI models and applicable terms may change. Updates will be reflected on this page.
Very useful
Not really
Very useful
Not really