Privacy Policy

Subtext ("we", "us", "our") is operated by Jameson Dog GmbH, registered in Munich, Germany. We take data privacy seriously — especially because the data you share with us is among the most personal information you have.

This policy explains what data we collect, why we collect it, how we process it, and your rights under the General Data Protection Regulation (GDPR) and applicable law.

1. What Data We Collect

Account Data

• Email address (if you sign in with email)
• Google or Apple account identifier (if you use social sign-in)
• Anonymous user ID (if you use the app without signing in)

Chat Export Data

• Chat and message exports you choose to bring in — from WhatsApp, Telegram, Google Chat, Gmail (read-only access to the one thread you pick), SMS (via a backup file you export yourself; a separately distributed sideload build can instead read the one conversation you pick directly, with your permission), and iMessage (via the Subtext Mac helper, which uploads only the one conversation you select)
• Participant names as they appear in the export
• Message timestamps, sender attribution, and message text
• Voice notes, if you choose a WhatsApp "with media" export — these are transcribed to text by an AI transcription service so they can be part of your analysis (see Section 4). Video files are discarded on your device and never uploaded

Survey Data

• Personality survey responses — communication style, love languages, attachment style, personality traits, and optionally self-reported neurodivergence and emotional patterns
• Relationship role quiz results (from the website quiz)

Some of those optional answers — self-reported neurodivergence or emotional patterns — can be special-category data under GDPR Art. 9. We process them only with your explicit consent, which you give when you start an analysis, and only to shape that analysis.

Payment Data

Purchases are handled entirely by third-party providers (Apple App Store, Google Play, RevenueCat, or Stripe). We receive only your purchase status, transaction identifiers and purchase timestamps. We never receive or store your card number, bank details, or any other payment credentials.

Usage Data

• Firebase Analytics events (aggregated; you can turn analytics off in the app under Settings → Privacy)
• Crash reports via Firebase Crashlytics (covered by the same toggle)
• Device type, OS version, app version
• Approximate location (country only) — derived once if the app shows you a therapy-referral option, from Google's infrastructure where possible, otherwise via a one-time IP lookup (see Section 4). We store only the country, never your IP or precise location.

2. How We Use Your Data

• Analysis: We process your chat exports using AI (Anthropic Claude) to generate relationship insights, communication profiles, and narrative reports. Voice notes in "with media" imports are transcribed to text by OpenAI's Whisper so they can be analyzed too.
• Improvement: Aggregated, anonymized analytics help us improve the product. We never use individual chat data for model training.
• Communication: We send transactional emails (report ready, account updates) and, only with consent, marketing communications.
• Safety: We run automated safety classification to detect indicators of harm. See Section 8.
• Compatibility and connection: If you opt in, we use insights about your communication style to power compatibility and connection features — including, where you choose, suggesting connections with other people. These features are off by default; you turn them on in the app, where we ask for your specific consent (including for any sensitive information) at that time, and you can withdraw it whenever you like.

Legal bases

Under the GDPR we must tell you the lawful basis for each purpose. Ours are:

Purpose	Legal basis
Providing the service — analysis, stories, the chatbot	Performance of contract (Art. 6(1)(b)); explicit consent for message content and any special-category survey answers (Art. 9(2)(a))
On-device analytics (volumes, sentiment, response times)	Performance of contract (Art. 6(1)(b))
Voice-note transcription ("with media" imports)	Performance of contract (Art. 6(1)(b)); the same explicit consent
AI-generated arc images (thematic descriptions only, no raw messages)	Performance of contract (Art. 6(1)(b))
Safety classification	The same explicit consent as the analysis it forms part of; legitimate interest (Art. 6(1)(f)) — safeguarding users
Push notifications	Consent (Art. 6(1)(a))
Waitlist and access emails	Performance of contract / consent (Art. 6(1)(a)-(b))
Therapy-referral routing (derived country)	Legitimate interest (Art. 6(1)(f))
Analytics and crash reporting	Consent (Art. 6(1)(a)) — off-switch in Settings → Privacy

You can withdraw consent at any time (Art. 7(3)); that doesn't affect processing already carried out on the basis of it.

3. How We Store and Protect Your Data

• Encryption: All data is encrypted in transit (TLS) and at rest (AES-256, Google-managed keys). Subtext is not end-to-end encrypted — our servers process your messages in order to analyze them; that's the product.
• Location: Our database layer is hosted in the EU. Our production infrastructure is being set up in the EU (Frankfurt region); during the current beta, uploads and processing run on Google Cloud infrastructure in the US. AI analysis is carried out by our processors in the US — see Section 4.
• Access: Your chat data is processed automatically. A small number of authorized operators can access limited samples and diagnostic data through an access-controlled internal console — solely for support you request, quality assurance, and abuse prevention, under confidentiality.
• Retention: We delete your raw chat export within 3 days of upload — immediately once your deep dive is written — unless you explicitly ask us to keep it (the "Help improve Subtext" toggle). What stays is your story: the insights, plus the sampled messages that power your in-app chat, until you delete it. You can clear the retained messages behind any single story (keeping its insights, turning off its chat), delete any story, or delete your entire account — all from inside the app, any time (see Section 6). One thing to know about the audio episode: it quotes messages word for word, and clearing the messages behind a story does not erase it. Deleting the story, or your account, does.

4. Third-Party Processing

• Anthropic (Claude AI): Your conversation text is sent to Anthropic's API for analysis, along with your survey context and chatbot messages — in full where it fits the model's context budget (the usual case), or as a stratified sample across the whole conversation where it doesn't. Anthropic does not use API data to train its models. Processed in the US.
• OpenAI (images + voice transcription): Thematic image descriptions (no raw messages) are sent to OpenAI to generate relationship arc images. If your WhatsApp export includes media, voice-note audio is sent to OpenAI's Whisper for transcription (videos are discarded on your device and never uploaded). OpenAI does not use API data to train its models. Processed in the US.
• ElevenLabs (podcast audio): To produce the audio episode about your story, the written episode script is sent to ElevenLabs for text-to-speech. That script is written by our AI from your deep dive and can contain participants' first names and messages quoted word for word. Processed in the US.
• Spotify: To resolve your story's soundtrack, we search Spotify with AI-written track title + artist queries — no personal data is sent.
• Postmark: Sends our transactional emails (waitlist, access) — receives your email address and the email content.
• RevenueCat: Manages in-app purchases — receives your user ID and purchase receipts. No card details ever reach us or RevenueCat beyond what the app stores process.
• ipwho.is: If the app resolves a therapy-referral option and Google's own geo header is unavailable, your IP address is sent for a one-time country lookup. Only the country is stored.
• Google Cloud Platform: Infrastructure hosting (Firestore, Cloud Storage, Cloud Functions).
• Firebase: Authentication, analytics, crash reporting, push notifications. Push notification text can include participants' first names — visible on your lock screen.

We do not sell, rent, or share your personal data with any third party for advertising, marketing, or data brokerage purposes.

5. International Data Transfers

Our database layer is in the EU. During the current beta, uploads and server-side processing also run in the United States on Google Cloud; our EU production infrastructure (Frankfurt) is being brought up. Some processing always involves a transfer to the United States:

• Anthropic (AI analysis) — we rely on Anthropic's Data Processing Addendum and the Standard Contractual Clauses (SCCs) approved by the European Commission.
• OpenAI (image generation, voice transcription) — OpenAI's DPA and SCCs.
• ElevenLabs (podcast audio) — SCCs.
• RevenueCat (purchase management) — SCCs.
• Postmark (transactional email) — SCCs.
• ipwho.is (IP-to-country lookup) — receives your IP address for a single lookup.
• Stripe (web payments, when available) — certified under applicable data transfer frameworks.

Only the minimum data necessary for each service is transferred.

6. Your Rights Under GDPR

As a data subject in the EU, you have the right to:

• Access: Request a copy of all personal data we hold about you (Article 15).
• Rectification: Correct inaccurate personal data (Article 16).
• Erasure: Request deletion of your data ("right to be forgotten") (Article 17). We will delete all account data, chat exports, reports, and derived insights. You can do this yourself in the app, or via subtext.love/delete-account.
• Restriction: Request that we limit how we process your data (Article 18).
• Portability: Receive your data in a structured, machine-readable format (Article 20).
• Objection: Object to processing based on legitimate interest (Article 21).
• Withdraw consent: Where processing is based on consent, withdraw at any time.

To exercise any of these rights, contact us at privacy@subtext.love. We will respond within 30 days.

You also have the right to lodge a complaint with a data protection supervisory authority. The authority responsible for us is the Bayerisches Landesamt für Datenschutzaufsicht (lda.bayern.de); you can also contact the authority of your own EU/EEA country.

7. Your Chat Partner's Data

When you upload a conversation you also share messages written by someone else. Subtext processes those messages to build your story.

By uploading a conversation, you confirm that:

• You were a participant in it;
• You have a lawful basis to process it — usually the GDPR household exemption for personal use, or your chat partner's consent;
• You won't use Subtext to harass, stalk, or harm anyone;
• You understand your chat partner's messages will be processed by the AI services described in Section 4.

We process chat-partner data on your instruction. We don't independently contact, profile, or identify chat partners; their names appear only inside your own private stories, are never used for marketing, and are never sold. If a chat partner asks us to delete their data from your uploads, we'll notify you and comply with valid requests under applicable law.

8. Content Safety

Subtext runs automated safety classification during chat processing to detect potential indicators of domestic violence, self-harm, coercive control, and child safety concerns. This processing is done in the legitimate interest of user safety.

When safety concerns are detected, we surface relevant helpline information to the user. We do not report to authorities or share this information with third parties. The safety classifier runs locally on our servers — no raw messages are shared for this purpose.

9. Cookies and Tracking

The Subtext website uses essential first-party storage (your quiz result, and a per-visit ID that links the pages of one visit together) and, only if you accept them in the cookie banner, Google Analytics 4 cookies plus two first-party items of our own: subtext_aid, a random ID with no name or email in it, and subtext_attr, which records the link or campaign that brought you here. Both last 180 days, are created only with your consent, and are deleted if you withdraw it. They are stored against your waitlist entry, so they are erased along with it and with your account.

We use campaign parameters on links we publish ourselves to see which of our own channels actually work. No advertising cookies, no third-party tracking pixels, and no cross-site tracking are used anywhere, and analytics is denied by default until you consent.

The public share page (share.subtext.love) sets no cookies and loads no third-party scripts. It counts views and store clicks on our own servers so we can tell whether sharing works, and its fonts are served from our own domain.

10. Children

Subtext is for adults only — you must be 18 or older to use it. We ask for your birth date once at sign-up, use it solely to confirm you're over 18, and don't store it. We do not knowingly collect data from minors; if we become aware that we have collected data from a user under 18, we will delete it promptly.

11. Additional Disclosures for US Users

California (CCPA/CPRA). If you're a California resident you also have the right to know what personal information we collect, use and disclose; the right to delete it; the right to opt out of its sale or sharing — we do not sell or share personal information for cross-context behavioural advertising; and the right not to be discriminated against for exercising these rights. Contact privacy@subtext.love.

Other US states. We comply with applicable state privacy laws including Virginia (VCDPA), Colorado (CPA) and Connecticut (CTDPA). The rights in Section 6 generally apply to all users regardless of location.

12. Changes to This Policy

We will notify registered users of material changes via email at least 30 days before they take effect. The current version is always available at this URL.

13. Contact

Jameson Dog GmbH
Klenzestr. 43, 80469 Munich, Germany
Privacy inquiries: privacy@subtext.love
Everything else: hello@subtext.love