Patentmania
Depatent

Privacy Policy

Last updated: 2026-11-12

Data controller

Agonist Development AB, Offerdalsgatan 3, 162 64 Vällingby, Sweden. Org.nr 559471-7849. Contact: support@patentomania.com.

Summary

Patentomania turns your voice and text into patent drafts. To do that, your audio and invention text leave your device and are processed by our servers and by the third-party speech and AI services listed below. We ask for your explicit consent to this before you use the app, we do not sell personal data, and we do not track you across other companies' apps or websites.

What we collect

  • Audio recordings you make in the app, sent to our servers for transcription.
  • Invention content: transcripts, typed text, drafts, claims and audit results you create.
  • Account data: an internal user ID; if you use Sign in with Apple, the name and email address Apple shares (possibly Apple's private relay address); if you use email and password, your email address and a salted bcrypt hash of the password.
  • Device identifier: a random device ID issued by our servers (not the Apple advertising identifier), used for push-notification preferences and session management.
  • Purchase data: subscription status and receipts, handled through Adapty and Apple.
  • Usage events and diagnostics: in-app events linked to your user ID (for example sign-in, draft completed), and crash reports provided by Apple.

Who receives your data and why

When you accept the consent screen, the following recipients may process your data. Which speech engine is used depends on your engine choice in Settings.

  • Patentomania servers (operated by Agonist Development AB, reached through Cloudflare Tunnel): your account, audio, invention text and drafts. Purpose: run the app.
  • Speech-to-text providers receive your voice audio (during live dictation, a continuous stream of audio) and return the transcript text. Our code can call these providers: Deepgram, AssemblyAI, Cloudflare (Workers AI), ElevenLabs, Speechmatics, Soniox, Gladia, Microsoft Azure (AI Speech), Mistral (Voxtral), Groq, Google (Cloud Speech-to-Text), and Amazon (Transcribe). Which one handles a given recording depends on your engine choice in Settings and on our fallback order (by default Deepgram, then AssemblyAI, then Cloudflare, then our own server). A self-hosted whisper.cpp server operated by us is the last fallback and shares nothing with third parties. We have no zero-retention agreement with any speech-to-text provider. Each provider's documented retention is listed in the section "Speech-to-text providers: what they keep" below.
  • AI model providers receive transcript and draft text to structure patent drafts and run audits: Cloudflare Workers AI, Groq, Google (Gemini), Cerebras, Mistral, Anthropic and OpenAI. They are called through a LiteLLM gateway operated by us.
  • Adapty (subscription management): purchase status, receipts and your user ID.
  • PostHog (analytics), self-hosted by us on a server in the Netherlands: usage events linked to your user ID, to improve the app. On this website it receives page views only if you accept analytics in the cookie banner. Requests go to ph.patentomania.com and pass through Cloudflare and a relay server in Germany before they reach the Netherlands server. No data goes to PostHog's own cloud.
  • Apple: Sign in with Apple, App Store payments, crash reports.
  • Stripe: payments made on our website only (not inside the iOS app). Stripe receives the payment details you enter there.

Claim-safe mode for unfiled inventions

For claims of inventions you have not yet filed, we prefer AI models whose provider documents that it does not retain submitted text (currently Cloudflare Workers AI). This is a preference, not a guarantee: we have no zero-retention agreement with any provider. Other text (for example transcripts used for interview questions or translation) may also use the remaining providers above, each under its own data-processing terms.

Some of these providers, for example Anthropic, may retain submitted text for a limited period (typically up to 30 days) for abuse monitoring under their standard terms; we have no zero-retention agreement with them. By their terms these providers do not train on API data, and we do not sell data. If confidentiality before filing matters to you (patent novelty, trade secrets), do not enter unfiled material that you cannot disclose to such providers. You can withdraw your consent at any time, and AI features will then be turned off.

Retention

  • Audio: the file on your phone is deleted after a successful upload. On our servers, audio is deleted 30 days after upload by default; it is kept longer (up to 365 days) only while an invention is in an active filing state, and you can opt out of retention for an individual disclosure, in which case it is purged after a 24-hour grace period.
  • Transcripts, drafts and account data: kept until you delete them or your account.
  • Speech and AI providers: retention is set by each provider's terms. For speech-to-text we have no zero-retention agreement with any provider and we do not currently call provider deletion endpoints, so audio and transcripts stay with the provider for as long as its own policy says (see below). Deleting your data on our side does not delete copies held by these providers; ask us and we will contact them on your behalf.
  • Analytics events: kept on our own server until your account is deleted.
  • Purchase records: kept as long as tax and accounting law requires.

Speech-to-text providers: what they keep

What is sent: your audio is sent from the app to our servers (HTTPS), and from our servers to the speech-to-text provider that handles the request. The provider returns transcript text, which we store with your invention. Live dictation streams audio to the provider while you speak. We have no zero-retention agreement with any of these providers. The retention below is what each provider documents; we link to the source, and it can change on the provider's side.

  • Deepgram: by default requests take part in its Model Improvement Program and data may be retained to improve its models; no retention period is published. Retention is limited to processing time only if the request sets mip_opt_out=true, which we do not currently set. Source
  • AssemblyAI: no formal zero-retention policy for pre-recorded audio; completed transcripts are stored for 30 days by default unless a shorter time-to-live is set or the transcript is deleted through its API; uploaded audio is removed within 24 to 48 hours. We have not configured a shorter time-to-live. Source
  • Cloudflare (Workers AI): documents that it does not retain inputs or outputs of standard inference after they are no longer needed, and does not use customer content to train models. Source
  • ElevenLabs: audio input and text output are retained by default; Zero Retention Mode exists only for select enterprise customers, and we are not one. Source
  • Speechmatics: audio, transcripts and job configuration of batch jobs are kept for 7 days, then deleted automatically. Source
  • Soniox: does not retain audio or transcripts for real-time requests; results of asynchronous requests are stored and deleted automatically after 30 days. Source
  • Gladia: retention depends on the plan; zero data retention is an Enterprise option that we do not have. Source
  • Microsoft Azure (AI Speech): does not retain data for real-time and fast transcription; for batch transcription the result is kept in Microsoft-managed storage (or in storage we specify) until a configured time-to-live expires or the job is deleted. Source
  • Mistral (Voxtral): inputs and outputs are kept for 30 rolling days for abuse monitoring unless zero data retention is approved for the account, which we have not requested. Source
  • Groq: documents that it does not retain inference inputs and outputs by default, with limited temporary logging in some cases, and offers a Zero Data Retention setting in its console. We have not verified that this setting is switched on for our account. Source
  • Google (Cloud Speech-to-Text): does not log customer audio or transcripts unless data logging is opted in for the project. Source
  • Amazon (Transcribe): may store and use content to provide and improve the service unless an opt-out is requested from AWS; transcription jobs can be deleted through its API. We have not requested an opt-out. Source

Delete your data

In the app you can delete an individual invention (with its drafts and audio) and you can delete your whole account (Settings, then Account, then “Delete account”; permanent deletion after a 30-day grace period). You can also opt out of audio retention for an individual invention, and email support@patentomania.com to request deletion. These actions delete the data on our servers. They do not yet delete copies held by speech-to-text providers, which follow the retention described above; on request we will contact the relevant provider to ask for deletion.

Withdraw consent

Open the app, go to Settings, then Legal, and tap “Withdraw AI data-sharing consent”. From then on no audio or invention text is sent to the services above and AI features are unavailable until you consent again. Data already processed is covered by the retention rules above, and you can ask us to delete it.

Delete your account

Open Settings, then Account, then “Delete account”. Your account and inventions are scheduled for permanent deletion after a 30-day grace period; signing in again within 30 days cancels the deletion. After 30 days we also purge the associated audio. Subscriptions are managed by Apple and must be cancelled separately in your Apple ID settings. You can also email support@patentomania.com.

International transfers

Several recipients are located in the United States and other countries outside the EU/EEA. Where required, transfers rely on a legal transfer mechanism such as the EU–US Data Privacy Framework or Standard Contractual Clauses; ask us for details.

Your rights (GDPR)

If you are in the EU/EEA or UK you have the right to access, correct, delete, restrict or object to processing of your data, to data portability, and to withdraw consent at any time without affecting earlier processing. Email support@patentomania.com and we will respond within one month. You may also complain to the Swedish Authority for Privacy Protection (IMY, imy.se) or your local supervisory authority.

Children

Patentomania is not directed to children under 16 and we do not knowingly collect their data.

Changes

If the list of recipients changes materially, we update this page and ask for your consent again in the app.

Contact

support@patentomania.com