Do Mote's AI providers train on or keep your data?
No. The AI providers Mote uses don't train their models on the recordings, writing and other content that students and teachers create in Mote. Most of them also delete that content as soon as a request is done, or within a short, fixed window.
How Mote sets this up
- Business terms only. Mote uses each provider's paid API or cloud service, never a consumer app.
- Opt-outs switched on. Some providers use customer data to improve their models unless you tell them not to. Mote has opted out with every one of them.
- Zero retention enforced on every request. When a request goes through OpenRouter, Mote requires providers with a zero data retention policy and excludes providers that collect data.
- Short retention where we can set it.
Provider by provider
| Provider | Trains on Mote data? | How long data is kept | Source |
|---|---|---|---|
| OpenRouter | No. OpenRouter doesn't use inputs or outputs for model training. | OpenRouter doesn't store prompts or responses. Mote's requests only go to providers with a zero data retention policy. | Zero data retention, Data collection, Privacy policy |
| Google Cloud (Gemini on Vertex AI, Speech-to-Text, Text-to-Speech, Translation) | No. Google doesn't use customer data to train or fine-tune AI models without permission. | Gemini: cached in memory for up to 24 hours. Prompts flagged by abuse detection may be logged for up to 90 days. Speech-to-Text: longer recordings produce a transcript that's stored for about 5 days. Text-to-Speech and Translation: not stored. | Service Specific Terms, section 18, Vertex AI data retention, Abuse monitoring, Speech-to-Text data usage, Text-to-Speech data logging, Translation data usage |
| Microsoft Azure (speech, language detection, and OpenAI models reached through OpenRouter) | No. Microsoft doesn't use customer data to train foundation models without permission. Prompts sent to OpenAI models on Azure aren't available to OpenAI. | Speech to text and text to speech: not stored. Language detection: text may be kept for up to 48 hours for debugging, then deleted. Language models: not stored, except content flagged by abuse monitoring, which may be kept for review. | Microsoft AI data commitments, Speech to text data privacy, Text to speech data privacy, Language data privacy, Azure OpenAI data privacy |
| Amazon Web Services (Anthropic and OpenAI models on Amazon Bedrock, Polly) | No. Bedrock inputs and outputs aren't used to train models and aren't shared with model providers. Mote's AWS organization is opted out of AWS using Polly content for service improvement. | Bedrock: not stored by default. Some models keep content flagged by abuse detection for up to 30 days. Polly: not kept for service improvement. | Bedrock FAQ, Bedrock abuse detection, AWS AI services opt-out |
| Cerebras | No. Cerebras's terms don't give it the right to use service content to train models. | Inputs and outputs aren't retained. | Terms of Use, Privacy Policy, section 6 |
| Anthropic | No. Anthropic may not train models on customer content from its API. | Deleted within 30 days. Content flagged as violating Anthropic's usage policy may be kept for up to 2 years. | Commercial Terms, Data retention |
| OpenAI | No. Data sent to the OpenAI API isn't used to train OpenAI models. | Abuse monitoring logs are kept for up to 30 days. | Data controls in the OpenAI platform |
| Deepgram | No. Mote opts every request out of Deepgram's Model Improvement Program. | Kept only for as long as needed to process the request. | Model Improvement Program, Terms of Service, section 3 |
| AssemblyAI | No. Mote's account is opted out of model training. | Audio and transcripts are deleted automatically after 1 day. AssemblyAI keeps some metadata for billing. | Data retention and model training, Data controls |
| ElevenLabs | No. Mote's account is opted out of ElevenLabs using its content for model training. | Used only for an optional seasonal voice. Generated speech is kept in Mote's ElevenLabs account history. | Is my data used to improve ElevenLabs models?, Terms of Use |
Caveats worth noting
- Abuse monitoring. Several providers keep a small amount of flagged content for longer so they can investigate misuse. That content still isn't used for training. The table gives each provider's stated limit.
- Providers change their terms. We update this article when a provider changes its terms or when we change providers.
Related articles
- Who do you share my data with? is the full sub-processor list.
- What AI services does Mote use, and who provides them?
- How long will my Mote content be kept available?
If your district needs a data processing agreement or more detail about a specific provider, email support@mote.com.
Provider terms last checked: September 2026.