Subprocessor and Provider List
Status: initial list, under legal review. Last updated: November 10, 2026.
Section 2.6 of the Terms of Service refers to this list. It names the third parties that receive personal data in Customer Data when Iterait provides the Services, and where they process it. Iterait processes this data in its role described in section 9.2 of the Terms.
All Iterait infrastructure runs outside Ecuador, in the United States. Customer data therefore leaves Ecuador when you use the Services.
Subprocessors
Iterait engages these subprocessors to perform the Services.
| Subprocessor | Data it receives | Where |
|---|---|---|
| Google Cloud (Firestore, Cloud Run, Cloud Storage, Logging, Cloud Tasks, Secret Manager) | All data stored and processed by the platform | United States |
| OpenAI (Responses, transcription, Realtime, files, vector stores) | Conversations, voice notes, call recordings, knowledge documents | United States |
| Google Gemini (Developer API) | Conversations, media, videos | Global |
| Anthropic | Conversations, for agents configured to use its models | United States or as chosen by the provider |
| Baseten | Prompts sent through an internal administration endpoint | United States or as chosen by the provider |
| OpenRouter and the model providers behind it (including OpenAI, Anthropic, DeepSeek, GLM and MiniMax models) | Conversations, for agents configured to use these models. Evaluation traffic for every agent: simulated test conversations and their grading, and the review of production conversations to find weak points | United States for OpenRouter. The model runs where its provider hosts it, which can include China |
| xAI | Call audio, customer photos | United States |
| ElevenLabs | Call audio, transcripts | United States |
| Jev (typesafe.ai) | Conversation transcripts | To be confirmed |
| Kling | Images and prompts | Singapore |
| Twilio | Phone numbers, call audio | United States |
| PostHog | Staff identifiers, organization names, pseudonymous participant identifiers, masked session replays | United States |
| Langfuse | Prompts and request metadata | To be confirmed |
| SendGrid, Gmail SMTP | Staff and customer email addresses and email content | United States |
| Pipedream | Customer contact details and connected-account tokens for Salesforce and Google Sheets | To be confirmed |
How long AI providers keep data
Section 8.5 of the Terms refers to this table. The retention below is what each provider publishes for its API, and it applies even when conversation storage is off.
| Provider | What Iterait requests | Provider retention |
|---|---|---|
| OpenAI | Conversations are sent with storage off (store: false). Knowledge files and vector stores stay until they are deleted. Embed agents are the exception: OpenAI stores their responses for 30 days and their agent state until it is deleted. Session retention is to be confirmed. | Abuse-monitoring logs for up to 30 days, longer where the law requires it or to protect its services or others from harm. Prompt caches for up to 24 hours on models that use extended caching |
| Google Gemini | Conversations are sent with storage off (store: false). | Prompts and outputs for 55 days, to detect policy violations |
| Anthropic | No conversation storage. | Inputs and outputs deleted within 30 days. Content flagged for policy violations is kept up to 2 years, and its classification scores up to 7 years |
| ElevenLabs | Keeps call records that Iterait reads back after a call. No retention setting is sent. | To be confirmed |
| OpenRouter and the model providers behind it | Every agent and evaluation request asks OpenRouter to use only providers that do not collect user data (data_collection: deny). They also ask for providers with zero data retention (zdr), except evaluation and Playground requests routed to OpenAI or Anthropic models. | To be confirmed. Requests routed to OpenAI or Anthropic keep those providers' retention, shown above |
| xAI, Baseten, Kling, Jev | No retention setting is sent. | To be confirmed |
Independent providers
These providers operate their own platforms under their own terms, as described in section 2.6 of the Terms. Payment providers act this way for certain processing.
| Provider | Data it receives | Where |
|---|---|---|
| Meta (Graph API, Conversions API, Pixel) | Messages, hashed email and phone, IP address, click identifiers, appointment events | United States, and data centers abroad while messages are processed |
| TikTok | Messages and account identifiers, when you connect a TikTok account | To be confirmed |
| X | Direct messages and account identifiers, when you connect an X account | To be confirmed |
| Payphone, DeUna, Stripe | National ID number (cédula), email, phone, card data, bank account details | Ecuador or United States |
Integrations you connect
When you connect one of these services, Iterait sends it customer contact details on your instruction. The service and its location are your choice.
Odoo, Zendesk, Salesforce and Google Sheets (both through Pipedream), Google Calendar, WooCommerce, Pentalab.
Data lookup sources
Some features look up debtors or prospects by national ID number or name in these sources.
| Source | Where |
|---|---|
| DataConsulting, EcuaBusqueda, Función Judicial | Ecuador |
| Google Places, Apollo | United States |
Questions
Questions about this list go to the contact in our Privacy Policy.