Privacy
Every AI meeting tool says it takes privacy seriously. Almost none of them can say your audio never leaves your device — because for most of them, sending your audio to their servers is how the product works. This page explains the difference between a privacy promise and a privacy architecture, and why that difference decides which tool you can actually use for the meetings that matter.
The Core Idea
A privacy policy tells you how a company intends to handle your data after it receives it. That intention can be sincere — and it can also change with an acquisition, a breach, a subpoena, or a quiet terms update. Privacy architecture works differently: if the transcription happens on your device, the vendor never possesses your audio at all. There is nothing to mishandle, nothing to retain, nothing to leak. You do not have to trust the promise, because the data never made the trip.
How It Works
Meeting Intelligence is a native Swift app that transcribes using Apple's Speech framework — the same on-device speech recognition technology built into iPhone, iPad, and Mac. Audio is captured, converted to text, and stored locally. Transcription is only half the story, though: AI analysis is where summaries, Decision Intelligence, and live guidance come from. Meeting Intelligence gives you a spectrum of options for that layer, so you choose exactly how private your pipeline is.
On Apple Silicon Macs, AI analysis can run entirely on the machine using local MLX models. Audio and transcript both stay home. Maximum privacy, no per-meeting cloud dependency.
The zero-setup path: transcript text (never audio) is analyzed by the managed Pro tier for live guidance and executive summaries. You decide whether that tradeoff fits the meeting.
Already have an Anthropic or OpenAI account with terms your organization has vetted? Use your own API key, and analysis runs under your agreement, not a third party's.
Run analysis against your own Ollama setup where configured — for teams that want local models on their own terms and hardware.
Who Needs This
For plenty of meetings, cloud processing is a perfectly reasonable tradeoff. But some professions do not get to make that tradeoff casually — and for them, "audio never leaves the device" is not a feature, it is the qualifying requirement.
Privileged conversations and a cloud transcription pipeline are an uncomfortable pairing. On-device processing keeps the privilege analysis simple.
When conversations touch sensitive personal information, the cleanest answer to "where does the audio go?" is "nowhere."
Your client's confidential information should not transit a third-party server because of your note-taking app. On-device transcription means it never does.
Material nonpublic information has no business in an unvetted pipeline. Local transcription — and fully local analysis on Apple Silicon — keeps the perimeter where it belongs.
Get Started
Privacy does not mean giving up capability. Meeting Intelligence still delivers live strategic questions with Ask This, real-time explanations with Quick Context, risk flags with Watch For, Decision Intelligence in every summary, and Meeting Memory that compounds across meetings — all on top of transcription that never leaves your device. Built by Edgepoint Solutions LLC.
Try Pro free for 7 days. No payment required to start. Transcript-only mode remains free.