The current versioned Belo AI subprocessor list for infrastructure, AI model providers, communications, authentication, billing, and observability.
Status: DRAFT — counsel review required (not operative until counsel-approved publish)
Why: who processes data for us (procurement / DPA).
When users see it: public page; referenced by DPA and trust page.
URL: 1
Translations: pt-BR · EN · es-ES · fr-FR (same path; app locale)
Source: 2026-07-21 reconciliation with lawyer commercial briefing
Counsel gate: qualified lawyer approval remains the final legal gate before operative publication.
Version ID: 2026-07-21-en.2
Changelog (2026-07-21-en.2): Restructured by provider function (function + permitted alternative). Added / made explicit: xAI (Grok), Cerebras, Exa Labs, Google Cloud Translation v3, Mailgun (Sinch), Cloud SQL for PostgreSQL, GA4 (consent-gated). Removed as listed subprocessors: AlloyDB, Typesense Cloud, Resend. Pre-authorization architecture (GCP ↔ AWS equivalents) and cloud-neutral BYOC via Order Form Supported Cloud Environment. New-subprocessor notice 15 days via List update (replaces prior 30-day List language).
Belo Inteligência Artificial Ltda, CNPJ 55.692.612/0001-84 (“Belo”) engages the following Subprocessors to help provide the Services (AI-assisted data analysis platform — see Terms of Service §1). Each Subprocessor is bound by written data-processing terms imposing obligations at least equivalent to those in the Belo Data Processing Agreement (where applicable).
This list is a versioned procurement artifact, organized by function (provider class), not by a single cloud vendor. Permitted alternatives for each function (e.g., Google Cloud Platform ↔ Amazon Web Services) are pre-authorized: by accepting this List, the customer authorizes use of any provider identified on it as the current provider or as a permitted alternative. Switching between pre-authorized providers takes effect by updating this List, without prior notice, provided the processing role, safeguards, and transfer mechanism remain equivalent. Advance publication of List updates (§4) and DPA objection rights apply only to the addition of new Subprocessors (not already listed), for paid customers.
New Subprocessors (not already listed) are announced by updating this List at least fifteen (15) days in advance. The published List update constitutes notice. Switching among providers already listed is reflected on this List without prior notice. Email alerts for List updates are opt-in via privacy@belo.ai or belo@belo.ai (subject: “Subprocessor notifications”).
Notes on AI inference. Belo routes generation requests to one or more of the providers above depending on customer model selections and entitlement tier. Prompts and outputs are processed under each provider’s data-processing / enterprise API terms. Belo does not train foundation models on Customer Content. Providers under enterprise API configurations used by Belo are instructed not to use customer inputs to train foundation models. Provider availability varies by plan and configuration; this list does not guarantee continuous availability of every model.
Platform search. First-party textual and similarity search on the platform runs against native PostgreSQL (it is not third-party subprocessing).
Supported Cloud Environment: the cloud environment for a BYOC deployment is as defined in the Order Form (or equivalent technical annex). Current: Google Cloud Platform (GCP only). When offered: Amazon Web Services (AWS). Microsoft, where applicable, appears only as a social login provider (Microsoft account / Entra ID) — not as a hosting cloud or BYOC target — unless the Order Form expressly states otherwise.
For BYOC customers:
See the Corporate/BYOC Addendum for support, update, and open-source boundaries.
Belo Inteligência Artificial Ltda
CNPJ: 55.692.612/0001-84
Rua da Consolação, 2302, Consolação, São Paulo/SP – CEP 01302-001 – Brazil
Email: belo@belo.ai / privacy@belo.ai
Belo AI
