Skip to main content

Models & Training Data

Harmony processes your meeting data with AI-powered transcription and summarization, while ensuring your conversations are never used to train external AI models or shared with others.

Zero Training Guarantee​

We never use your data to train external AI models.

When you record a meeting with Harmony:

  1. Your data is not used to train OpenAI, Google, or Anthropic models.
  2. Your data is not shared with other customers.
  3. Your data is not sold to third parties.

How we process your data​

We use top-tier AI providers (like OpenAI and Google) to generate transcripts and summaries, but we operate under strict Zero Data Retention (ZDR) agreements.

  1. Encryption: Audio/text is sent to the model provider via an encrypted connection.
  2. Processing: The model generates the transcript or summary in real-time.
  3. Deletion: The provider deletes the data immediately after processing. They do not save a copy.

Training Harmony's internal models​

Harmony develops some internal, specialized models (e.g., for speaker identification).

  • These are trained on anonymous, opt-in data only.
  • By default, your workspace is opted out of training.
  • You can choose to opt-in to help improve accuracy, but this is completely voluntary.

Self-hosted intelligence​

For organizations with extreme security requirements (Defense, Government, Finance), we offer Self-Hosted Intelligence.

  • Deploy Harmony on your own AWS/Azure/GCP cloud.
  • Connect your own LLM infrastructure (e.g., local Llama 3 or private Azure OpenAI).
  • No data ever leaves your controlled environment.

Deleting your data​

You retain full ownership of your data.

  • Soft Delete: When you delete a meeting, it is removed from the UI immediately.
  • Hard Delete: Backups and archives are purged within 30 days.
  • Workspace Deletion: If you cancel your account, all data is permanently erased.