Privacy Policy
Last updated: 19 October 2026
In short
- Fermo’s servers keep your account email, your credit balance, usage counts and, if you add a passkey, its public key. They do not store audio, transcripts, screen content or your notes.
- Your history, notes and files stay in your browser, on your device. Your history and notes also go to your Google Drive if you turn on backup. You can delete them at any time.
- To work, call audio goes to a transcription service (Deepgram). Your questions, passages from your notes and files, and the text and a picture of the part of your screen you share go to an AI service (OpenAI, with Anthropic as a backup). If you add files, Cloudflare’s AI service also gets their passages and your questions, to search the files; it keeps no copy of the text. If you turn on follow-up drafts, the transcript goes to the AI service once more, after the call.
- Fermo listens only after you start a session and the people present agree.
- We do not sell your data, show ads, or use your content to train AI models.
1. Who we are
Fermo is a web app. It listens to a conversation that you choose, writes it down and suggests talking points. It runs at fermo.app and app.fermo.app.
Voidraft LLC runs Fermo. Voidraft LLC is a Delaware limited liability company with the mailing address 2108 N St #18080, Sacramento, CA 95816. In this policy, “we” and “us” mean Voidraft LLC. “You” means a person who uses Fermo.
For questions and requests, email [email protected].
This policy tells you what data Fermo uses, where it goes, how long we keep it, and what you can do. Our Terms of Service give the rules for using Fermo.
2. How Fermo uses data in a session
A session starts only when you press Start and confirm that the people present agree (section 8). Then:
- Your browser captures the call audio that you choose to share with Fermo, and your microphone.
- The audio goes through Fermo’s server to a transcription service. The service sends back text. No audio is recorded.
- If you added files on Sources, Fermo looks for the passages that fit each question. It sends the question and the two lines before it to Cloudflare’s AI service, also while the person is still speaking. The service sends back a search vector (a list of numbers), and your browser compares it with the vectors of your passages.
- When a question comes, Fermo sends the question, recent lines of the transcript, passages from your notes and files, and the text and a picture of the part of your screen that you share to an AI service. The AI service sends back suggested points.
- Your browser shows the points, with the names of the files that each point drew on. It saves the transcript and a report on your device, and in your Google Drive if you turn on backup (section 6).
Files (optional). When you add a file on Sources (PDF, Word, text or Markdown), your browser reads it and cuts its text into passages. Fermo sends each passage once to Cloudflare’s AI service to make its search vector, and again only when the file changes. The text, the passages and the vectors stay in your browser. Fermo does not keep the original file.
Follow-up drafts (optional). This is off by default. If you turn on Draft follow-ups after a session in Settings and press Draft a thank-you note and action items on a report, Fermo sends the session transcript to the AI service once more, after the call. The AI service sends back a draft. Fermo keeps the draft with the report: on your device, and in your Google Drive if you turn on backup.
Fermo’s server passes this content along in memory. It does not store the content and does not write it to logs.
3. What Fermo’s servers keep
Our servers keep only what we need to run your account and your credit, and to stop abuse.
| What | Why | How long |
|---|---|---|
| Your email address | To sign you in and to write to you about your account | Until you delete your account |
| Your Google account ID, if you sign in with Google. We do not keep the name on your Google account | To sign you in with Google | Until you delete your account |
| Sign-in codes and sign-in sessions, kept only in scrambled (hashed) form | To sign you in and to keep you signed in | Codes: until used, or 10 minutes. Sessions: until you sign out or they expire |
| If you add a passkey: its public key and its ID, a sign-in counter, and the name of the passkey provider, for example iCloud Keychain. The private key stays on your device or in your password manager. Fermo never receives it | To sign you in with your passkey | Until you remove the passkey in Settings, or delete your account |
| Your credit: balance, top-ups, and the minutes used in each session | To let you start live sessions | Until you delete your account |
| Purchase records: Stripe order IDs, amount, currency, billing country | For refunds, disputes and tax | 7 years. When you delete your account, we remove your email from them |
| Usage counts per day: minutes of speech service, number of answers, AI tokens, cost | To charge sessions from your credit and to answer billing questions | 13 months |
| Abuse counters, for example sessions started per hour, or passages prepared for search per day | To stop overuse and attacks | Until the counter resets |
| A flag that a session started after you confirmed that everyone agreed. It holds no names and no content | To show that sessions start only after the consent step | 13 months |
| Error logs (codes and request IDs, no content) and timing numbers (for example, how long an answer took) | To find and fix faults | Error logs: up to 7 days. Timing numbers: 90 days |
| Emails that you send to support | To answer you | 12 months after we close your request |
Our servers never store audio, transcripts, questions, answers, screen text, screen pictures, your notes, your files, their passages or search vectors, your Google Drive backup files or your Google access tokens.
Our database keeps restore points for up to 30 days. So when we delete data, it also leaves those restore points within 30 days.
4. What stays on your device
Fermo saves your content in your browser’s storage, on your device. We cannot see it. It includes:
- Session history: transcripts, reports and consent records. A consent record lists the names that you entered, who agreed, and when. It holds no transcript text.
- Fermo notes: for example About me, a fact sheet, interview notes and a project library.
- Files: the text of each file that you add on Sources, cut into passages, with a search vector for each passage. Fermo does not keep the original file. In Chrome and Edge, Fermo also keeps your browser’s permission to read the file again, so that it sees when the file changes. It checks every few minutes while a Fermo tab is open. Your browser can ask you again before Fermo reads a file.
- For each session: the names of the files that its answers drew on, and the moments when the talk turned to your files (a short quote of the line, and the file names).
- Settings.
If you turn on Google Drive backup, Fermo also saves each session (transcript and report), your notes and your follow-up drafts in your Google Drive (section 6). Your files are not backed up.
Fermo keeps session history by default. To stop this, turn on Don’t keep history in Settings. Fermo then deletes each new session when you close its report.
What you can do:
- Export all (Settings) downloads one file with your sessions and notes. Import restores it, in this browser or another one. Your files are not in it: add them again in the other browser. Export on a session saves its report, then its transcript.
- Remove on a file in Sources deletes its text, passages and search vectors from this device.
- Delete all (Settings) removes all sessions and their consent records from this device. Your notes, your files, your settings and any Google Drive backup stay until you delete them.
- Delete one session in History.
- Reset this browser (Settings) removes everything that Fermo saved in this browser.
- Clear the site data for app.fermo.app in your browser. This removes everything that Fermo saved on that device.
Please note:
- Each browser keeps its own copy. A new device starts empty. If you turn on Google Drive backup on it, Fermo restores your history from your Drive. When you delete your account in Settings, Fermo also clears its data on that device. The data on your other devices and in your Google Drive stays. Delete it on each device and in your Drive.
- Anyone who can use your browser profile can see this data. Your device’s login and disk encryption protect it. On a shared computer, turn on Don’t keep history.
- A browser can remove stored site data when the disk is full. Export what you want to keep, or turn on Google Drive backup.
Reminders (optional). This is off by default. If you turn on Remind me 5 minutes before a session in Settings, your browser first asks you to allow notifications. The open Fermo tab then shows a browser notification 5 minutes before a session that you planned in Fermo. Fermo’s server sends no notifications and keeps nothing for them.
Installing Fermo as an app. You can install Fermo from your browser, so that it opens in its own window from the Dock or the Home Screen. Installing gives Fermo no extra data. Fermo asks your browser to keep its storage, so that the browser is less likely to clear it when the disk is full. A small app file (a service worker) keeps a copy of Fermo’s own page files, so that the app opens fast. It never keeps your content or replies from Fermo’s server.
5. Who receives data, and why
We use the service providers below. Each one gets only what its task needs. The consent notice that Fermo shows before each session names them too.
Deepgram: speech to text
- Receives: the call audio that you share, your microphone audio, and the one short line that you say in the microphone check during setup. With the audio, Fermo can send a short list of key words, such as technical terms, to help the service spell them correctly.
- Why: to turn speech into text.
- Speaker labels: in a call, the service can label who is speaking (“Speaker 1”, “Speaker 2”). These labels do not identify a person.
- Training and retention: Deepgram can use audio to improve its models unless the customer opts out. Fermo opts out on every request. With the opt-out, Deepgram keeps audio only as long as it needs to process it.
OpenAI: answers, and backup speech to text
- Receives: questions, recent lines of the transcript, passages from your notes and files, and, on turns that read the screen, the text and a cropped picture of the part of the screen that you share. If you use follow-up drafts, it also receives the session transcript once more, after the call. If Deepgram is not available, your audio can go to OpenAI’s speech-to-text service instead.
- Fact sheet: when you press Make the fact sheet in Sources, OpenAI receives your resume and notes once, to write a short fact sheet. Your answers then use the sheet, not the raw text.
- Why: to write suggested points, the fact sheet and follow-up drafts, and to turn speech into text when Deepgram is down.
- Training and retention: OpenAI says that it does not train its models on data sent through its API unless the customer opts in. Fermo does not opt in, and Fermo turns off stored responses. OpenAI can keep messages for up to 30 days for abuse checks, even with stored responses off, and longer if the law requires it or to protect against harm.
Anthropic: backup answers
- Receives: only when OpenAI fails on a turn: the same content that OpenAI would get for that turn, including the screen picture.
- Why: so that you still get a point when OpenAI is down.
- Training and retention: Anthropic says that, by default, it does not train its models on inputs and outputs from its API. Fermo does not opt in. Anthropic deletes inputs and outputs within 30 days. If its safety systems flag a message as a breach of its usage policy, Anthropic can keep that message for up to 2 years, and the safety scores about it for up to 7 years.
Cloudflare: hosting, servers, search and email
- Receives: all traffic to fermo.app and app.fermo.app, including the content that passes through Fermo’s server in a session. Cloudflare also sees your IP address and browser details, to deliver the site and to block attacks.
- Search (Workers AI): only if you add files on Sources. Each passage of a file goes once to Cloudflare’s AI service, Workers AI, to make its search vector. During a session, each question and the two lines before it go there too, also while the person is still speaking, to find the passages that fit. Cloudflare keeps no copy of the text after the vector returns; it can keep usage numbers, such as how much text it processed. Cloudflare says that it does not use this content to train models. Fermo keeps the vectors in your browser. The model is Google’s EmbeddingGemma, in Beta on Workers AI. Google’s Gemma Terms of Use govern it.
- Bot check and email: Cloudflare Turnstile checks the sign-in page for bots. Cloudflare’s email service sends your sign-in codes. Cloudflare Email Routing forwards mail sent to [email protected].
- Why: to run Fermo’s website, app, server and database. Our database is encrypted at rest.
Stripe and Link: payments
- Receives: what you enter at checkout, such as your name, email, billing address and payment details. Fermo never sees your full card number.
- Role: Stripe’s Managed Payments service handles the payment. Sold through Link, LLC (“Link”), a Stripe company, is the merchant of record for your purchase. Link charges you, handles sales tax, sends your receipt and handles payment support. Voidraft LLC is the seller: it sells you the credit and provides Fermo.
- Stripe uses this data under its own privacy policy. You can ask Link to delete your payment data.
Google: sign-in and Drive backup
- Sign-in: if you choose Continue with Google, Google tells Fermo your email address, your name and a Google account ID. Fermo asks only for your basic profile and email. Fermo keeps only your email address and the account ID. It does not store your name.
- Drive backup: only if you turn it on. Google then stores your backup files in your Google Drive. See section 6.
We also disclose data when the law requires it, to protect people or Fermo from harm or fraud, or as part of a sale or merger of the business. After a sale or merger, this policy still applies to your data.
6. Google Drive backup
Google Drive backup (optional)
Backup is off by default. To turn it on, use Back up to Google Drive in Settings.
- What Fermo saves: one file for each session, with its transcript and report, that you can read in Drive. Also your Fermo notes and, if you make them, your follow-up drafts.
- Where: a folder named Fermo in your own Google Drive. Fermo only adds files. It never overwrites a file, so two devices do not clash. Fermo’s servers do not store these files.
- When: after each session, while a Fermo tab is open.
- Permission: when you turn on backup, Google asks you to give Fermo the
drive.filepermission. This is a separate step from sign-in. With this narrow permission, Fermo can see and change only the files that it created. It cannot see your other files. - Never backed up: your settings and your files. A session in the backup names the files that its answers drew on. Fermo records no audio, so there is no audio to back up.
- A new device: when you turn on backup there, Fermo restores your history from the Fermo folder.
- Turn it off: turn off Back up to Google Drive in Settings. Fermo then stops new backups. The files already in your Drive stay until you delete them. To delete them, delete the Fermo folder in Google Drive. To remove Fermo’s access, remove Fermo from the third-party connections in your Google Account.
Limited Use. Fermo’s use and transfer to any other app of information received from Google APIs will adhere to the Google API Services User Data Policy, including the Limited Use requirements. This statement covers Google sign-in and Google Drive backup.
This means:
- We use Google data only to provide the features that you see in Fermo.
- We transfer Google data only to provide those features (for example, to the AI service for an answer), for security, to follow the law, or in a merger or sale after you give your explicit consent.
- We do not use Google data for ads, and we do not sell it.
- People at Fermo do not read Google data, unless you agree, for security, or to follow the law.
- We do not use Google data to create, train or improve any AI or machine-learning model.
7. Voice data
Fermo makes no voiceprints. It does not keep a voice profile of you or of anyone else.
- During a session, the transcription service tells voices apart, so that Fermo can label lines (“Speaker 2”) and tell you when a new voice joins. Fermo keeps these voice groups in memory and forgets them when the session ends.
- We never sell, lease or trade voice data, and we never profit from it.
8. Consent, and the people you talk to
- Fermo does not listen until you start a session and confirm that the people present agree. There is one exception: in the microphone check during setup, you say one short line, and only that line goes to the speech service.
- Before each session, Fermo shows a notice that you can read out or paste in the chat. Copy notice copies it in one click. Start waits until you mark that each listed person agreed. When someone joins later, Fermo shows you a note, so that you can ask them.
- Fermo does not hide from screen sharing. Chrome shows its own sharing bar. If you share your whole screen in the call, other people can see Fermo’s answers.
- You are responsible for telling people and for getting the agreement that the law requires. See our Terms of Service.
If you talk with someone who uses Fermo. While their session runs:
- Your voice goes to the transcription service. No audio is recorded.
- Your words appear in the transcript. The transcript stays on the user’s device, and in the user’s Google Drive if they turn on backup. Passages go to the AI service when a question needs them.
- If the user added their own files to Fermo, your questions and the lines before them also go to Cloudflare’s AI service, to find passages in those files. Cloudflare keeps no copy of the text.
- Your name is kept on the user’s device only if the user typed it in the consent list.
- Anything on a screen that the user shares with Fermo can go to the AI service, as text and as a picture.
- If the user turns on follow-up drafts, the transcript goes to the AI service once more after the call, to draft a thank-you note and action items.
- Fermo’s servers keep nothing about you.
You can ask the user to stop Fermo at any time. The transcript is on their device, and in their Google Drive if they use backup, so ask them to delete it. You can also write to us at [email protected]. Our page for the people you talk to is fermo.app/about-fermo.
9. What we do not do
- We do not sell personal data. We do not share it for advertising. Fermo shows no ads.
- We do not use your content to train AI models.
- The app has no analytics tools, ad pixels or tracking scripts. The website fermo.app uses Cloudflare Web Analytics to count visits. It records page views, the page, the site that sent you, your country, your device type, browser and operating system, and how fast the page loads. It uses no cookies and does not track you across websites.
- We do not track you across other websites. So “Do Not Track” and Global Privacy Control signals change nothing: we already treat every visitor as opted out.
10. Cookies and browser storage
- Fermo uses one cookie to keep you signed in. The app needs it to work.
- Cloudflare’s bot check on the sign-in page can set a cookie that it needs for security.
- Fermo uses no advertising or analytics cookies.
- Fermo uses browser storage (section 4) only for the features that you use.
- If you install Fermo as an app, its service worker keeps a copy of Fermo’s own page files. It holds none of your content (section 4).
11. Security
- All traffic uses HTTPS.
- Our database is encrypted at rest.
- Provider keys stay on our servers. They are never in your browser.
- Fermo has no passwords. You sign in with an email code, with Google, or with a passkey.
- A passkey’s private key never leaves your device or your password manager. Fermo’s server keeps only the public key, which cannot sign you in by itself.
- Our servers do not log content.
- No system is perfectly secure. If a breach affects your data, we tell you as the law requires, and we tell the authorities where the law requires it.
12. Your choices and rights
Everyone. You can ask us to show, correct, export or delete the data that our servers keep about you. Email [email protected] from the email address of your account. We answer within 30 days. We can ask you to confirm the request from that address. We cannot see or delete the data on your devices or in your Google Drive. Use Export and Delete in the app, and delete the Fermo folder in your Drive.
Delete your account. Use Delete account in Settings, and press it again to confirm. Fermo deletes your account and clears its data on that device. Your Fermo folder in Google Drive stays in your Drive: delete it there. You can also email us. We delete your account data from our servers within 30 days. We keep the purchase records that tax law requires, without your email. To delete the payment data that Stripe and Link keep, ask Link.
California. If you live in California, you can ask to know, correct and delete the personal information that we keep. You can use an authorized agent. We do not sell or share personal information as California law defines those words, and we did not do so in the last 12 months. We do not treat you differently when you use your rights.
European Economic Area, United Kingdom and Switzerland.
- Voidraft LLC is the controller of your personal data.
- Legal bases: a contract with you (your account, your credit and the sessions that you start); our legitimate interests (security, stopping abuse and keeping Fermo working); your consent (Google Drive backup); and legal duties (tax records). You can withdraw consent at any time.
- Your rights: access, correction, deletion, restriction, objection, data portability and withdrawal of consent. You can also complain to your data protection authority.
- Transfers: we process data in the United States. Our providers process it in the United States and other countries. For these transfers, we rely on their data processing terms.
Other places. If your local law gives you similar rights, email us and we will help.
13. Where we process data
We run Fermo from the United States. Your data is processed in the United States and wherever our providers run their services. Fermo is in English and is open in every country. We may not be able to sell credit in every country.
14. Age
Fermo is for people 18 and older. It is not for children. If we learn that a person under 18 gave us personal data, we delete it.
15. Changes to this policy
When we change this policy, we post the new version on this page with a new “Last updated” date. If a change is important, we email account holders 14 days before it takes effect.
16. Contact
Voidraft LLC
2108 N St #18080
Sacramento, CA 95816
[email protected]