Does TrackHeros train AI on my meetings?

No. TrackHeros does not train AI models on your meetings, and it does not let its AI providers do so either. This answer is for anyone recording client calls on Google Meet and Microsoft Teams who needs to tell a client, or a client's security team, exactly where the recordings go and what happens to them there. The general questions to put to any notetaker are in questions to ask an AI notetaker about client data.

Which AI providers process my meetings?

Every model call except building the search index goes through OpenRouter, which passes it on to the company that runs the model. Every request requires a provider that keeps no copy of the content and refuses any use of it for training; a provider that cannot meet both is never used.

Step	Provider	Terms
Transcription of English-only meetings from the recording; summaries and minutes of meetings not held in English; the other steps that write the notes	Google's Gemini models on Google Cloud's Vertex AI, through OpenRouter	No copy of the content kept and no use of it for training, required on every request
Summaries and minutes of meetings held in English; answers when you ask about your meetings	Anthropic's Claude models on Google Cloud's Vertex AI, or on Amazon Bedrock when Vertex AI cannot serve them, through OpenRouter	The same requirement, on every request
Transcription of meetings held in a language other than English, or in more than one, and their live transcript during the call	A Whisper model run by DeepInfra, Together AI or Groq, whichever OpenRouter chooses	Terms that exclude training and keep no copy of the audio once it is transcribed
The search index over your meetings	Google Cloud's Vertex AI, directly rather than through OpenRouter	Enterprise terms that exclude training on your content

Dictating a question to the assistant goes through the same pipeline, comes back as text in the box you type in, and is not stored.

The processors that handle data for TrackHeros, including those that have nothing to do with AI, and what each one does, are described in the privacy policy.

Why does it matter which terms a provider uses?

Because "no training" is only as good as the contract behind it. The same company can offer a consumer AI app and an enterprise cloud service under very different terms, and a notetaker's promise depends on which of the two it calls. TrackHeros calls the cloud services, Google Cloud's Vertex AI and Amazon Bedrock, not a consumer app, and it reaches them through OpenRouter with the same two requirements on every request: no copy of the content kept, and no use of it for training. A provider that cannot meet both is never sent the request, so the no-training position does not rest on a consumer app's settings.

The same test applies to DeepInfra, Together AI and Groq, which transcribe meetings held in other languages; OpenRouter chooses among them, and all three exclude training on your content and keep no copy of the audio after transcription.

Does "no training" mean nothing is stored?

No. TrackHeros stores your recordings, transcripts, summaries and action items so you can read, search and ask about them. Storing them for you is the product; training a model on them would be a different use, and that is the one that does not happen.

They stay in the workspace of the client they belong to, are kept for as long as you keep that workspace, and the assistant inside a client reads only that client's meetings. How the workspaces are kept apart is on client workspaces that never mix. When you delete a client or your account, the data goes from live systems at once and from backups within 30 days; what happens to recordings when you delete a client has the detail.

Where can I check this in writing?

In three places, which say the same thing at different lengths:

• The four privacy statements on the homepage, which are the policy in short.
• The privacy policy, under "AI processing and model training" and "Who else processes your data".
• The about page, under what TrackHeros deliberately does not do.

If a client of yours needs more than that, such as the current processing regions for their meetings, write to TrackHeros and it will tell you.

Does this cover the meeting platform's own AI features?

No. Google Meet and Microsoft Teams have their own transcription and AI features, and if a client's organisation uses them, those are governed by the client's agreement with Google or Microsoft, not by TrackHeros. The TrackHeros bot joins as a separate participant and its recordings follow TrackHeros' terms only.