Speak AI uses the sub-processors below to deliver the service. We never train models on customer content: customer content is sent to AI providers only to process your request and return the result, and those providers are used on no-training API terms.
Infrastructure
| Sub-processor | Purpose | Processing location |
|---|---|---|
| Amazon Web Services | Hosting, file storage, AWS Transcribe | Canada (ca-central-1); certain workloads in us-east-1 |
| MongoDB Atlas | Database | Canada Central |
AI, transcription and translation
| Sub-processor | Purpose | Processing location |
|---|---|---|
| OpenAI | Large language model and speech, on no-training API terms | USA |
| Anthropic | Large language model, no-training | USA |
| Google Gemini | Large language model and translation, no-training | USA |
| Deepgram | Transcription | USA |
| AssemblyAI | Transcription | USA |
| Microsoft Azure Speech | Transcription | USA |
Voice agents
These apply only if you use Speak voice agents.
| Sub-processor | Purpose | Processing location |
|---|---|---|
| LiveKit | Voice agent | USA |
| ElevenLabs | Voice agent | USA |
| Twilio | Voice agent | USA |
Changes to this list
Customers under a data processing agreement (DPA) receive 30 days’ notice of new sub-processors. For questions, contact success@speakai.co.
Related: Third-party data privacy policy · Security