# Third-party subprocessors (/core/govern/security-model/third-party-subprocessors) **Target Audience**: Security, legal, and procurement teams reviewing external data processing for Toucan AI. ### TL;DR [#tldr] Toucan AI may rely on external providers **depending on your deployment and enabled features**. The table below summarizes common categories. Your contract and DPA define the authoritative subprocessor list for SaaS. *** ### When to use this [#when-to-use-this] Use this page for privacy assessments, vendor reviews, and understanding which external parties may process data when you use Toucan AI. *** ### Subprocessor overview [#subprocessor-overview] | Provider category | Purpose | Data that may be processed | When used | | ------------------------- | ---------------------------------------------------------- | --------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------- | | **LLM provider** | AI inference (chat, chart generation, metadata enrichment) | Prompts, conversation context, query-related context | When AI features are enabled | | **Product analytics** | Usage metrics, product improvement | Events, performance data; during beta, may include conversation-related telemetry | When analytics is enabled | | **AI tracing** | Debug and quality of AI workflows | AI conversations: questions, answers and the data excerpts the assistant read | As the organization chooses, see [which conversations are shared](/faq-and-troubleshooting/ai#conversation-sharing) | | **Billing** | Subscriptions and payments | Organization and billing identifiers | When billing is enabled | | **CRM** | Customer success and onboarding | Account email, organization profile fields | When CRM sync is enabled | | **OAuth providers** | Sign-in with Google, GitHub, LinkedIn | OAuth tokens and profile data | When social login is enabled | | **Email (SMTP)** | Transactional email | Email addresses, message content | When email is sent | | **Notifications (Slack)** | Internal alerts (e.g. signups, feedback) | Notification payloads | When webhooks are configured | #### LLM providers [#llm-providers] Toucan AI sends data to an LLM provider to run AI features. The default integration path uses **Anthropic** (Claude). Organizations can select a different provider and model from **Settings → AI behavior** (see [Choose your AI provider and model](/core/govern/ai-control-and-optimization/select-your-ai-provider)); the provider you select is the one that receives your prompts and context. **Typical data sent**: user messages, system instructions, database metadata snippets, and context needed to answer or build a chart. #### Analytics & tracing [#analytics--tracing] * **Product analytics** may record feature usage and AI-related events. * **Tracing tools** may record steps of AI execution for debugging and quality. #### Business & identity services [#business--identity-services] * **Billing** processes subscription data when billing is active. * **CRM** may sync signup and profile fields when enabled. * **OAuth** providers handle authentication when users sign in with a social account. *** ### Your responsibilities [#your-responsibilities] * Review the subprocessor list in your **DPA** or trust documentation for the authoritative legal register. * Disable or configure optional integrations according to your policy * Minimize personal data in tokens and AI context before it reaches any external provider. *** ### Related pages [#related-pages] * [PII & personal data](/core/govern/security-model/pii-and-personal-data) * [AI assistant data handling](/core/govern/security-model/ai-assistant-data-handling) * [Data storage & retention](/core/govern/security-model/data-storage-and-retention)