Slide 3 of 3

Privacy and security

Review how Assistant handles your data

Grafana Assistant processes your prompts and conversation context using third-party AI providers. Requests route through Grafana Cloud and never go directly from your browser to third parties.

Grafana Labs makes three commitments:

  • Data ownership: You retain ownership of your data.
  • No training on your data: Grafana Labs never uses your organization’s data to train AI models or for marketing.
  • No third-party retention: Providers process your prompts without storing or training on them.

Review the providers Assistant uses

Assistant uses different providers depending on the feature and the type of request.

FeatureProvidersProvider data retentionTraining on your data
ChatAnthropic, Google Vertex AI, or Amazon BedrockZero retentionNo
Web search (used on demand)TavilyZero-day retentionNo

Zero-day retention means Tavily doesn’t retain search terms after the day they’re processed.

Grafana selects the provider based on infrastructure configuration. You can’t choose which provider processes your requests.

Configure RBAC and access controls

The Assistant respects your Grafana permissions. If you don’t have access to a dashboard or data source, the Assistant can’t access it either.

Administrators can use RBAC to control who can open Assistant by scoping plugins.app:access to plugins:id:grafana-assistant-app. Refer to Manage Assistant access with RBAC for instructions.

Review when data leaves your environment

Assistant is not enabled by default. You don’t send data to any AI provider until you actively use an Assistant feature. Enabling the plugin in the Plugins catalog doesn’t send any data.

For full details on data handling, encryption, and access controls, refer to Manage your data privacy and security.

Script

Grafana Assistant processes your prompts and conversation context using third-party AI providers. Before any data leaves your environment, it’s worth understanding what gets sent, who processes it, and what protections apply.

The Assistant uses providers including Anthropic, Google Vertex AI, and Amazon Bedrock. None of these providers store your data for training purposes. Your prompts and query context are processed to generate responses and then discarded by the provider. Grafana Cloud retains conversation history so the Assistant can stay in context across a thread, but it never uses your organization’s data to train AI models.

The Assistant also respects your Grafana permissions. If you don’t have access to a dashboard or data source, the Assistant can’t access it either. Assistant is not enabled by default, so you don’t send data to any AI provider until a user actively uses an Assistant feature.

For the complete details on data handling, encryption, and access controls, refer to the privacy and security documentation.