Security

The data-handling story

AsYouSpeak only handles two kinds of data: the words spoken on your calls, and the specs and prototypes derived from them. Here's the full path that data takes through the product — nothing more, nothing hidden.

Transcripts are transient

Audio is transcribed in your browser, or you paste a transcript in yourself. The transcript text is sent to Anthropicto extract the spec and build the prototype. We don't retain raw call audio, and we don't keep the live transcript as a permanent record — it lives only long enough to produce the spec and prototype.

What is stored, and where

For paid users, the artifacts that power the Library are stored with managed providers rather than on infrastructure we hand-roll:

• ·Vercel Blob — the generated HTML prototypes, served over HTTPS.
• ·Upstash Redis — the spec, metadata, and your usage and budget counters.

Free-tier sessions are not persisted to a Library. Traffic to and from the app runs over HTTPS.

Access control

Paid features are gated behind an access code tied to an email address. Your Library and usage are scoped to your account, and we use your email to authenticate you. We don't expose another user's prototypes or specs to you.

Spend is enforced server-side

Budget caps, per-meeting limits, and build debouncing are enforced on the server, not in the browser, so they can't be bypassed by tampering with the client. The app also surfaces the live compute cost during a session so there are no surprises.

What we never do

We don't sell your data, and we don't share your transcripts, specs, or prototypes with anyone outside the sub-processors named above — and then only to run the product. For the full picture, see our Privacy page.

Reporting an issue

Found a vulnerability? Email security@asyouspeak.com and we'll get back to you.