Privacy Policy

Rhetoric is operated by Cortis LLC (Wyoming, USA). This policy says what we collect, why, who processes it for us, and the controls you have. The short version: your speeches stay yours, and Rhetoric never stores your raw audio.

What we collect

• Account data.Your email address and a password hash, managed by our authentication provider. We don't ask for your name unless you add one.
• Your content.Speeches you write, teleprompter transcripts, practice metrics (pace, coverage, recoveries, scores), and debate records. This is the product. It's stored so you can come back to it.
• Billing data.Handled by Stripe. Your card number never touches Rhetoric's servers; we store your subscription status and Stripe customer reference.
• Usage events. Product analytics (which features are used, where people get stuck) tied to your account id, never your speech text. You can turn this off on the cookie preferences panel.
• Support requests. If you ask for help or report a problem from Help & support, we keep your message, your reply address, and, only if you choose to attach it, a small technical summary you can preview before sending (browser and system family, app release, which screen you came from). It never includes your speech text, transcripts, or audio. We use it to reply to you, debug the problem, and prevent abuse of the form.

What we deliberately don't collect

Raw audio is never stored by Rhetoric. Voice tracking converts your speech to text live; what we keep is the transcript and the measurements, both visible to you. Where the audio goes depends on the recognition engine, which we label before a session starts:

• Your browser's own speech service(the public demo, and the fallback when the enhanced engine is not available): audio is processed transiently by that service (for example Chrome's or Safari's, which may process audio on their vendor's servers). Rhetoric does not receive it.
• The enhanced engine on a paid plan: your browser sends microphone audio directly to our speech provider (Deepgram) for transient processing. Rhetoric does not receive it.
• The free trial: the audio travels through a relay that Rhetoric operates (hosted on Supabase) on its way to the same provider, so that a session can be ended at its time limit. The relay passes the audio on; it does not record or keep it.

In every case, Rhetoric does not record or retain the audio itself.

How we use your data

• To provide the product: storing, syncing, and analyzing your speeches and sessions.
• AI features (analysis, coaching, the debate opponent and judge) send the relevant textto Anthropic's API. Per Anthropic's commercial terms, API data is not used to train their models.
• To bill you, send transactional email (account activation, billing notices), and answer support.
• To watch product health: aggregated usage, error reports with request ids, not speech content.
• We do not sell your data or run third-party ads. Your content is shared only with the people you invite to a speech, and only for that speech; no one else can see it.

Who processes data for us

• Supabase: database, authentication, and the relay that carries trial audio in transit (the relay does not store audio).
• Stripe: payments, subscriptions, and sales tax.
• Anthropic: AI text processing for the features above.
• Deepgram: enhanced speech recognition and live-voice debate. Sessions that use it are labeled before they start; audio is processed live to produce the transcript and is not stored by Rhetoric.
• Cloudflare: live collaboration relay. When you and your collaborators edit a speech together, the document text passes through a Cloudflare Worker and is held in memory and in encrypted object storage during and between live sessions; the durable copy of every speech stays in Supabase.
• Copyleaks: plagiarism scanning, only when you run a plagiarism check on a speech; the speech text is submitted for that scan and nothing else.
• PostHog: product analytics (optional, your call).
• Resend: transactional and support email.
• Vercel: hosting.
• Sentry: error monitoring.

Retention & your controls

• Your content stays while your account exists. From Settingsyou can export everything you've made, delete individual sessions (including transcripts), or delete the whole account. Deletion removes your content from the live database.
• If your subscription lapses, your library remains readable and exportable. We never hold your writing hostage.
• Billing records are kept as long as tax law requires.
• Support requests are included in your Settings export and are deleted with your account. Independently of that, closed requests are deleted within 12 months of being closed (rejected spam within 30 days). Replies exchanged over email follow the same schedule in our support mailbox.
• Free trial identity records. To prevent repeat trials, Rhetoric stores three keyed hashes (HMAC-SHA256 under a secret key) for each trial that was started: one derived from the email address, one from the device (a cookie), and one from the network address (for IPv6, the /64 prefix) at the time of sign-up. A network hash counts against new trials for 30 days. A hash cannot be turned back into the address, device or network it came from; it can only be compared with a new sign-up. The hashes are kept after the account is deleted, because otherwise deleting an account would earn a new trial. This retention serves the legitimate interest of preventing abuse of a free offer. To ask about these records, write to support@rhetoricspeech.com.

Cookies

We use a small set of cookies: session, a 5-minute subscription cache, and optional analytics. They're itemized on the cookies page, where the analytics opt-out lives.

Security

Data is encrypted in transit, access is enforced row-by-row in the database (your rows are readable only by your account), and API keys for AI services never reach the browser. No system is perfect; if a breach affects you, we'll tell you promptly and plainly. If you believe you've found a security vulnerability, please write to security@rhetoricspeech.com rather than the ordinary support form.

Children

Rhetoric is not directed at children under 16, and we don't knowingly collect their data.

Changes & contact

If this policy changes materially, we'll note it here with a new date and, for significant changes, email you. Questions or requests (access, export, deletion): support@rhetoricspeech.com, or use Help & support.