Privacy Policy

Last updated 2026-06-09.

This is the short version: we process public tweets you click Read this debate on. Everything else is opt-in or opt-out from a toggle in Settings. The long version follows.

1. What we collect

• Account data: email address (used to send you sign-in links), tier, account creation timestamp.
• Analyses: the public tweet content you ask us to analyse, the AI-generated debate map, and metadata (model version, latency, cost).
• Interaction signals: thumbs up/down, comments you leave on AI output, edits to drafts you copy, panel-open / panel-dwell timings.
• Optional surveys: occasional one-tap questions ("did this change your mind on the topic?") that you can disable at any time.

2. Why we collect it

• To deliver the analysis you requested.
• To save your briefs to your own History so you can re-read them (see History & storage, below).
• To improve the model and prompts, if you've left training on (see Training Data, below).
• To enforce per-tier quotas and detect abuse.
• To contact you about your account or, if you've opted in, product updates.

3. History & storage

When you're signed in we store the briefs you run (the AI-generated debate map and basic metadata) so you can re-open them in your History inside the extension. This is a feature of the service you asked for, so it applies whether or not you allow training. You can delete this data at any time via the DELETE /me endpoint or by emailing us.

The original harvested thread contents are only retained when you have training enabled (below); if training is off, we keep your brief for your History but discard the raw harvested thread.

4. Training data

If — and only if — you leave the "Use my analyses to help train a better Receipts" toggle on, your debate inputs (tweet text), our outputs (analyses, drafts), and the interaction signals above may be used to:

• Refine the prompts we send to language models (today).
• Fine-tune our own future models (later).

This consent is recorded per analysis at the time you run it, and it is independent of your access to the product — turning training off does not remove any features (your History still works). You can opt out at any time from the Settings panel inside the extension. Opting out applies prospectively — analyses already collected with your prior consent remain in our training corpus, but you can request deletion via the DELETE /me endpoint or by emailing us.

5. Early access

Receipts is in a limited early cohort: up to 100 accounts. Everyone gets 15 thread briefs per day (resets midnight UTC). Paid tiers and promotional discounts are not offered during this phase; we may add them later with notice by email.

New sign-ups have training-data and analytics consent defaulted to ON so we can learn from real usage; flip them off any time in Settings.

6. Third parties

• Anthropic processes your tweet content to generate analyses. We use Anthropic's API with prompt caching enabled. Anthropic privacy policy.
• Tavily processes specific factual claims for web verification. Tavily privacy policy.
• Resend sends transactional sign-in emails. Resend privacy policy.
• Stripe (when payments launch) handles billing.
• Hosting on Render or equivalent EU/US infrastructure.

7. Your rights

• Access: GET /me/data returns a JSON dump of everything we have on you.
• Deletion: DELETE /me soft-deletes your account and detaches your email from analytics rows. Email hello@loalab.art for hard deletion.
• Portability: the JSON export is machine-readable.
• Objection / withdrawal of consent: Settings → consent toggles.
• Complaints: contact your local data protection authority (e.g. ICO in the UK, CNIL in France).

8. Retention

Account rows are kept while the account is active and for 12 months after deletion to handle billing reconciliation. Training-corpus rows keyed only by hashed user ID are kept indefinitely unless deletion is requested.

9. Children

Receipts is not directed at children under 16 and we do not knowingly collect their data.

10. Changes

Material changes will be notified via the email address on file.

11. Contact

hello@loalab.art