AI Transparency

Effective Date: 04 July 2026 · Last Updated: 04 July 2026

MeetMind AI uses artificial intelligence to transcribe meeting recordings, generate summaries, extract action items, and answer follow-up questions. This page explains how AI is used, what models power the service, and how your data flows through the system.

1. AI Models Used

  • Whisper-large-v3-turbo (via Groq): a high-speed, open-weights speech-to-text model used for standard fast audio transcription with precise timestamping.
  • Deepgram Nova-3: an advanced speech recognition model used in Meeting Mode for multi-speaker identification and diarization.
  • Llama 3.3 70B (via Groq / NVIDIA NIM): an advanced open-weights large language model used for generating executive meeting summaries, extracting key decisions and action items, and powering the interactive meeting chat assistant. Transcript text (not audio) is passed to generate structured responses.

2. Data Flow

Here is how your data moves through the system:

You upload an audio or video file
audio compressed & preprocessed to 16kHz mono
Speech recognition engine transcribes audio into timestamped text
transcript text sent to Llama 3.3 70B inference engine
Llama 3.3 extracts executive summary, decisions, and action items
structured results encrypted in your Supabase account
Results displayed in dashboard for review, chat, and export

3. AI Limitations

AI-generated outputs have inherent limitations:

  • Transcripts may contain errors, especially with accents, background noise, overlapping speakers, or technical terminology.
  • Summaries, action items, and decisions are AI-interpreted and may miss context, misattribute statements, or include inaccuracies (sometimes called "hallucinations").
  • Chat responses are based on the transcript and summary, and may not accurately reflect what was discussed.

4. Human Verification

We strongly recommend that you review and verify all AI-generated outputs before using them for any decision-making, reporting, or professional purpose. MeetMind AI does not include human review or moderation of outputs. All results are generated entirely by AI.

5. Data and Model Training

Your audio, transcripts, summaries, and chat messages are not used to train or fine-tune any AI model. Your data is used solely to provide the service to you.

6. Use Restrictions

MeetMind AI is designed for meeting analysis and productivity. It is not intended for, and should not be relied upon for, medical diagnosis, legal advice, financial decisions, or any other professional advice requiring certified expertise.

7. Reporting Issues

If you encounter an incorrect, misleading, or harmful AI output, please report it to meetmindai.help@zohomail.in. Include the meeting ID and a description of the issue so we can investigate.

8. Responsible AI

We are committed to using AI responsibly. We select models and providers that align with responsible AI practices, and we continuously evaluate the quality and safety of AI outputs in our service.