Privacy Policy
Effective November 15, 2026
Steno is built to know as little about you as possible. Recording and transcription run on your device. There's no account, no tracking, and we don't sell any data.
Audio and transcripts
Audio is recorded and transcribed on your phone (the Parakeet and Whisper models run offline). Audio never leaves your device – not to iCloud, not to our server – unless you share or export it yourself. It isn't part of iCloud or computer backups either, so it isn't restored to a new phone from a backup; transcripts, summaries and tasks are (through iCloud sync).
Speech models
The models for transcription and speaker recognition are downloaded once from Hugging Face (huggingface.co). Like any website, Hugging Face sees your device's IP address during the download; nothing else about you or your recordings is sent. The models aren't part of backups – they can be downloaded again at any time.
Voice recognition
When "Remember voices" is on, Steno stores a voice fingerprint of each speaker with every recording and, for voices you name, their averaged fingerprint. Fingerprints stay on this phone only: they don't go to our server, to iCloud or into backups. You delete a remembered voice in Settings → Known voices; the fingerprints stored with a recording are deleted with it, and turning "Remember voices" off deletes all fingerprints stored with recordings.
iCloud sync
When it's on, transcripts, summaries, tasks, recording settings and – for a recording linked to a meeting – the event title and invitees' names are stored in your private iCloud (Apple CloudKit). Only you can access them; we can't. You can turn it off in Settings → iCloud.
AI summaries and chat (Steno AI)
Without your consent Steno sends no transcript or other content (you give it in the app and can withdraw it in Settings → AI summaries and tasks). Even without consent, the app only checks your plan (or subscription) and that the install is genuine (App Attest) with the Steno server – nothing from your recordings. Once you allow it, to write a summary and find tasks, it sends the Steno server the transcript text with speaker names, the recording date, the chosen template, the recording length and – if the recording is linked to a calendar meeting – the event title and invitees' names. For "Ask" (chat) it sends your question and the earlier messages of the conversation, the recording's title, date, summary and transcript; for questions about your whole library, the titles, dates, summaries, key points and tasks of all recordings plus the full transcripts of at most three that fit the question. The server passes this to OpenAI and returns the answer to the app.
- The Steno server runs on Cloudflare. It doesn't store or log transcripts or answers, and request logging (Workers Logs) is turned off. As the network provider, Cloudflare processes your device's IP address to deliver the request and protect the service from attacks; we don't store IP addresses.
- OpenAI doesn't use API data to train its models. To prevent abuse it may keep it for up to 30 days and then deletes it. We send requests with an instruction not to store responses.
- With every request OpenAI also receives a pseudonymous identifier (a "safety identifier") – a one-way fingerprint of your install's or subscription's identifier. It contains no name or email; it lets OpenAI detect abuse without blocking every Steno user. It is the same Steno AI ID you find in the app under Settings → Contact the developer.
On the Steno server we keep only the technical data needed to run and protect the service:
- your install's App Attest identifier and public key (proving a request comes from the genuine Steno app),
- counters: how many free summaries you've used this month, what processing cost this month and how many requests were made in the last hour,
- for free summaries, a one-way fingerprint (hash) of the transcript that the transcript can't be recovered from – so sending the same recording again, for example after a lost connection, doesn't use up another summary,
- for a subscription, the App Store transaction identifier, product and expiry date (to check the subscription is active).
This data isn't linked to your name, email or Apple Account, but it is kept under your install's or subscription's identifier (pseudonymously). Transcript fingerprints are deleted after a day, hourly counters after 2 days, monthly ones after at most 13 months. The App Attest key and the subscription record are kept while you use Steno AI; we delete them on request.
Your own API key
In Settings → AI → Advanced you can use your own Claude (Anthropic), ChatGPT (OpenAI) or Gemini (Google) key instead of Steno AI. By checking the key (or confirming in the app) you agree that the transcript text (with the same details as described above) goes directly to that provider – not through our server – under the terms you have with them. With a free Gemini key, Google may use the content to improve its services. The key is stored only in your device's Keychain.
Instead of these providers you can also enter another service with an OpenAI-compatible API – for example your own server running an AI model on a computer at home, on your local network or over a VPN. By verifying the connection (or confirming in the app) you agree that the transcript text (with the same details as described above) goes directly to the server you entered – not through our server – under the terms of that service or whoever runs it. The consent covers only that server; if you enter a different one, the app asks again.
Steno Plus subscription
Payments are handled by Apple. We never see payment details. Apple sends us only signed transaction data (identifier, product, dates) and notices of renewals or refunds.
Other apps
Tasks and notes go to Reminders, Calendar, Things, Todoist, Notion or Obsidian only when you turn that on or send them. Todoist and Notion receive them directly with your token; we don't see them.
Calendar: the event title and invitees' names are read on the device, saved with the recording (and synced with it through iCloud) and sent with the transcript when a summary is made. When you ask, Steno adds a task to your calendar as an event.
Recording other people
You are responsible for being allowed to record the people you talk to. Where the law requires it, you need their consent – the simplest way is to tell everyone that you're recording.
Diagnostics
Steno uses no analytics or tracking. The diagnostic log (technical details only – never audio, transcripts, tasks or keys) stays on your phone and reaches us only if you attach it to an email.
Children
Steno isn't directed at children, and we don't knowingly collect any data from them.
Who processes the data
The controller is Jan Šetina, developer of Steno, contact gusto.satori-3e@icloud.com. Processors: Cloudflare (server hosting), OpenAI (AI processing), Apple (iCloud, App Attest and payments). Some of them process data outside the EU, with appropriate safeguards under the GDPR. The legal basis is performance of a contract – providing the features you ask for. Abuse prevention and rate limiting (the install identifier, the counters and the pseudonymous identifier sent to OpenAI) rest on our legitimate interest in running the service securely and fairly.
Your rights
You delete recordings right in the app (which also removes them from iCloud). You have the right to access, rectify, erase, restrict and port your data, the right to object to processing based on legitimate interest, and the right to lodge a complaint with a data protection authority. For data on our server, write to gusto.satori-3e@icloud.com and include your Steno AI ID (Settings → Contact the developer) – that's how we find it.
Changes
If this policy changes, we'll update the date above and mention significant changes in the app.