Privacy Policy

Version 2 · in effect from 22 October 2026

This policy explains what we collect when you use our services, what we do with it, who else handles it, and how long we keep it. The short version: we collect what we need to run your account and read your documents, we keep what we learn about forms and de-identified data about their contents, never who a document is about, we don't sell personal data or your documents, and nobody trains AI models on your documents.

1. Who we are

Our services are run by ClaimPartner Operations LLC, a Delaware limited liability company, 8 The Grn Ste A, Dover, DE 19901 ("we", "us"). This policy covers the websites, apps and APIs we offer under our Terms of Service: today extract_insurance, at extractinsurance.dev and its subdomains, and any name or address that replaces them ("the services").

Questions about privacy: hello@extractinsurance.dev.

2. Two kinds of data, two roles

• Your account data: who you are and how you use the services (section 3). We decide how it is used, so for this data we are the controller.
• Your documents and the results we return (section 4). You decide what to send and what it's for; we process it to provide the services to you, on your instructions. For this data we act as your processor (or "service provider"). The exception is the data we derive (section 4), which identifies no one; for that we are the controller. Insurance documents often name other people, such as policyholders, claimants and drivers. You are responsible for having the right to send us their information and for any notices or consents the law requires.

3. Account data we collect

• Who you are: your email address, your name if you give it, your company name, your role in an account (owner or member), and the accounts you belong to.
• Signing in: if you sign in with GitHub or Google, we take only your verified email address from them. We don't keep their access token or any other profile information. If you sign in with an email link, the link is single-use and expires after 15 minutes.
• Your team: the email addresses you invite, who invited them, and when.
• Usage: the jobs you run, how many documents were read, what they cost, and when; and how you and your AI agents use the API, the command-line tool and the MCP server (which calls, from which tool, how long they took), by account and key. API keys are stored only as a one-way hash; we can't see the key itself.
• Analytics: how the website and app are used: pages viewed, clicks, where you came from and where you stopped. We connect your visits over time with an id kept in your browser (section 10), and may record sessions to see where people get stuck. In the app, recordings hide all text, so no document content or result is recorded.
• What you tell us: votes and comments on upcoming features, and corrections and notes you add on the review page.
• Technical data: we use your IP address briefly, in memory, to limit repeated sign-in attempts; we don't store it. Our hosting provider keeps its own request logs. Our application logs record only ids, timings, counts and statuses, never document text, values or quotes.

We use account data to run your account, sign you in, send the emails you ask for (sign-in links and invites), keep your usage and bills correct, keep the services secure, and improve them, including through analytics. We don't use it for advertising.

4. Your documents and results

How they're protected

Every connection uses TLS 1.2 or newer, both to us and from us to the companies in section 5. Documents are encrypted (AES-256-GCM) before they are stored, on disks that are themselves encrypted. Each account's documents, results, workflows and keys belong to that account only.

How long we keep them

• Documents sent through the API are deleted as soon as their job finishes, unless you choose to keep them for 1 to 90 days.
• Documents you upload in the app (Try a document, workflow samples) are kept for your retention period: 30 days unless you set 1 to 90. Samples stay while they are attached to a workflow.
• Results are kept for 30 days unless you set 1 to 90.
• After that they are deleted. What remains is your billing record (counts and amounts, no content) and the data we derive, below.

Data we derive

To provide and improve our services, we may keep information about the structure of documents, de-identified data about their contents, and usage data. De-identified data doesn't identify you, your clients or the businesses and people your documents are about. We don't try to re-identify it, and anyone we let use it must agree not to either. Benchmarks and statistics show only combined figures that don't single out any business or customer.

No training on your documents

We don't train AI models on your documents or results, and the model providers we use don't either: their API terms rule it out.

The improvements setting

Each account has an improvements setting, on by default, which an owner can turn off at any time in the account settings. While it is on:

• we may read a small share of your documents a second time, in the background and at our cost, to test new models and tune how fields are read. From those reads we keep which fields and document parts were found, not the values. The documents are still deleted on your normal schedule;
• corrections you make on the review page change your own results. We may add a correction to our internal test set only after we have checked it; the test set is used to measure accuracy, not to train models.

When it is off, your documents aren't read again after their job. The data we derive, above, doesn't depend on this setting.

5. Companies that handle data for us

We use these companies (subprocessors) to run the services. None of them may use your data for their own purposes.

Under their API terms, the model providers may keep what we send them for a limited time (up to 30 days) to detect abuse, and then delete it. The website loads its fonts from Google Fonts, so your browser shares your IP address with Google when you visit it. Before we start using a new company, we add it here and email account owners at least 14 days ahead. If you object, tell us; if we can't address it, you can close your account and we refund your unused credits.

We don't sell personal data and don't share it for advertising. We disclose data to others only to run the services as described here, when the law requires it, or to protect our users and the services. If our business or a service is sold or merged, its data passes to the buyer or successor, under this policy.

6. Where data is processed

We store data in the United States. As section 5 shows, some processing happens in the European Union, and a model provider may process data in other countries. Some countries, including those of the EU and the UK, restrict sending personal data abroad. If you need a data processing agreement with the EU standard contractual clauses for your use of the services, write to hello@extractinsurance.dev.

7. Security

Besides the measures in section 4, sign-in links, invites and API keys are stored only as hashes, sign-in is rate-limited, and our host is audited to SOC 2 Type II. No system is perfectly secure; if a breach affects your data, we will tell you as the law requires.

8. How long we keep account data

We keep account data while your account exists. If you ask us to delete your account, we delete its data within 30 days, except records we must keep by law, such as billing records.

9. Your rights

Depending on where you live, you can ask us to access, correct, delete or export your personal data, or object to or restrict how we use it. Write to hello@extractinsurance.dev; we answer within 30 days. If you are in the EU or UK, you can also complain to your data protection authority. California residents: we don't sell or share personal information, and we won't treat you differently for using these rights.

If your information is in a document one of our customers sent us, please contact that customer; we will help them answer you.

10. Cookies

We use two cookies, both to sign you in: extract_session keeps you signed in, for 30 days; extract_oauth protects a GitHub or Google sign-in while it is in progress, for 10 minutes. We don't use cookies for advertising or to track you across other sites.

For analytics, PostHog keeps a random id in your browser's storage, not in a cookie, to connect your visits across pages and sessions. Clearing this site's data in your browser removes it.

11. Children

The services are for businesses and aren't meant for anyone under 18.

12. Changes

Each version of this policy has a number and the date it takes effect, and we keep every version. If we change it, we post the new version here; if a change matters, we email account owners before it takes effect.