Legal
Subprocessors
Last updated: June 18, 2026
To deliver our AI voice-agent services, AIforge Tech Private Limited (operating as Finn) engages the third-party subprocessors listed below. Each is bound by data-protection obligations no less protective than those in our Data Processing Addendum, and processes personal data only to provide the function described. Zero-data-retention configurations noted below apply where the provider offers them and the Customer’s plan includes them; on all plans, providers are bound by no-training API terms.
Recordings, transcripts, and stored data stay at rest in the workspace region the Customer selects (Iowa, Mumbai, Singapore, Sydney, London, or Frankfurt). Live speech and language processing routes to each vendor’s endpoints in the regions listed below — select a workspace region to see the processing locations that apply to it.
A given call’s exact data flow depends on two choices the Customer makes: the workspace region, and the voice and language-model stack configured for each agent. Each call touches only the vendors in that agent’s configuration — the table below is the complete set a configuration can draw from, not the set every call uses.
| Subprocessor | Category | Purpose | Data location |
|---|---|---|---|
| Google Cloud Platform | Infrastructure · AI | Cloud hosting, storage & compute, speech services, and language models (Gemini, and Anthropic Claude via Vertex AI) · no-training API terms; zero data retention where offered | Data at rest & speech: customer workspace region (Iowa, Mumbai, Singapore, Sydney, London, or Frankfurt) · model inference: US / EU / APAC Vertex endpoints (model-dependent) |
| Microsoft Azure | Infrastructure · AI | Cloud hosting, storage & compute, speech services, and language models (OpenAI models via Azure OpenAI) · no-training API terms; zero data retention where offered | Data at rest & speech: customer workspace region · model inference: US / EU / APAC data zones per deployment type (realtime models process globally) |
| OpenAI models (via Microsoft Azure) | AI services | Language and realtime speech models for live calls, served through Azure OpenAI — call data is processed by Microsoft, not OpenAI, and is never used for training | Azure deployment regions: US / EU / APAC per deployment type; realtime models process globally |
| Anthropic Claude (via Google Cloud Vertex AI) | AI services | Language models for live calls, served through Vertex AI — call data is processed by Google Cloud, not Anthropic, and is never used for training | Vertex AI endpoints: United States / EU / Singapore, or global |
| Deepgram | AI services | Speech-to-text for live calls · no-training API terms; zero data retention where offered | United States (EU and Australia endpoints available) |
| ElevenLabs | AI services | Text-to-speech for live calls · no-training API terms; zero data retention where offered | United States (EU, India, and Singapore residency available on Enterprise) |
| Cartesia | AI services | Text-to-speech for live calls · no-training API terms; zero data retention where offered | United States / EU / APAC (routed by origin) |
| Sarvam AI | AI services | Text-to-speech for live calls · no-training API terms; zero data retention where offered | India / United States |
| Inworld | AI services | Text-to-speech for live calls · no-training API terms; zero data retention where offered | United States (enterprise/on-prem deployment options) |
| Smallest AI | AI services | Text-to-speech for live calls · no-training API terms; zero data retention where offered | India / United States (geo-routed) |
| Twilio | Telephony | Call & SMS origination and termination | United States (EU and Australia regions available) |
| Plivo | Telephony | Call & SMS origination and termination | United States |
| ContactOut | Data enrichment | Contact data enrichment for Customer-uploaded audiences | United States |
| Dodo Payments | Payments | International payment processing | Global |
| Lemon Squeezy | Payments | International payment processing (merchant of record) | Global |
| Resend | Communications | Transactional & contact-form email delivery | United States |
| Intercom | Support | In-app support messaging | United States |
| PostHog | Analytics | Product-usage analytics | United States or EU (Frankfurt) |
| Sentry | Monitoring | Error & performance monitoring | United States |
| Razorpay | Payments | Subscription & payment processing | India |
| Cloudflare | Security | Bot protection & CAPTCHA (Turnstile) | Global |
The specific AI providers that process a given call depend on the voice and language-model configuration the Customer selects for each agent. Twilio and Plivo act as Finn’s sub-processors for call origination and termination. The downstream telecommunications networks they route calls over act as mere conduits for transmission — they do not store personal data beyond what is transient to carrying the call — and are not Finn sub-processors. Any Customer-connected integrations (e.g. CRM or calendar accounts) are governed by the Customer’s own agreements with those providers.
The specific processing location for any subprocessor is available on request at [email protected].
Where a subprocessor processes personal data outside the workspace region selected by a Customer, that transfer is carried out under the Standard Contractual Clauses (or an equivalent safeguard) incorporated into our Data Processing Addendum.
Changes to this list
We update this page when we add or remove a subprocessor. As set out in our Data Processing Addendum, we provide at least fourteen (14) days’ advance notice of a new subprocessor; customers may subscribe to that notice by contacting us and may object to a new subprocessor within that period.
Contact
Questions about our subprocessors? Reach us at [email protected].
Get started
Hire Finn and scale with confidence.
Move from idea to live voice automation — securely, reliably, and without operational risk.
Fort-nightly Launches
We move quickly and get you what you need
Powerful Tools
Pre-built dashboards, reports, automations, more