Last updated: 13 October 2026
This page lists the third parties that receive the text you submit to Lenz, or data derived from it, such as the claims read from it and the search queries written from it. It applies to the website, the API, the connector for AI assistants and WhatsApp. The first list below is the providers that receive the text itself; the second is the providers that receive only something written from it; the third is account data.
Text you submit
These providers receive the text you submit, or text that can contain it. A single check does not reach all of them: SendGrid and WaSenderAPI apply only to checks you submit on the website or WhatsApp, so a check submitted through the API never reaches them.
| Provider | What it receives | Purpose | How long it keeps it1 | Location |
|---|---|---|---|---|
| Google Cloud (Google LLC) | Everything Lenz stores and processes: submitted text, results, Word documents you upload, account data and logs | Hosting, database, file storage (a private bucket for uploaded documents), task queues, caching and logs | Lenz sets how long it is stored — see the Privacy Policy. Google’s own deletion and backup schedules then apply on top | United States |
| Anthropic | Submitted text and text derived from it: claims, search queries, source passages and follow-up questions | AI processing for claim analysis and fact-checking | Up to 30 days, and up to 2 years for text Anthropic’s systems flag as breaking its usage policy | Stored in the United States; processed in the US, Europe, Asia or Australia |
| Google (Gemini, via Vertex AI) | Submitted text and text derived from it: claims, search queries, source passages and follow-up questions | AI processing for claim analysis and fact-checking | None of the text Lenz sends is stored. For an account set to zero retention, Gemini runs with zero data retention: a grounded check searches through Web Grounding for Enterprise, which logs nothing. For other accounts, the queries and the context sent with them to Google Search when a check is grounded: up to 3 days, not linked to Lenz or to you | Any Google Cloud region worldwide |
| OpenAI | Submitted text and text derived from it: claims, search queries, source passages and follow-up questions | AI processing for claim analysis and fact-checking | Up to 30 days, for abuse monitoring, and longer where the law requires it | United States |
| Perplexity | Submitted text and text derived from it: claims, search queries, source passages and follow-up questions | AI processing for claim analysis and fact-checking | Perplexity stores each request and its reply. Not used for an account set to zero retention | North America |
| xAI | Submitted text and text derived from it: claims, search queries, source passages and follow-up questions | AI processing for claim analysis and fact-checking | Not kept: Lenz’s xAI account runs with zero data retention, so the text Lenz sends and the replies are never written to disk. xAI records the request’s timing, token count and model, for billing, and no part of the text | Global |
| SendGrid (Twilio) | The claim, its summary and answers to your follow-up questions, in emails about your checks | Sending email; used only for checks you submit on the website or WhatsApp | Not yet confirmed | United States |
| WaSenderAPI | Messages you send to and receive from the Lenz WhatsApp number, and your phone number | Connecting the Lenz WhatsApp number; used only if you use Lenz on WhatsApp | Not yet confirmed | Not yet confirmed |
¹ “How long it keeps it” is the provider’s own period: how long the text may stay with them after they have answered. How long Lenz keeps it is a separate question, and the Privacy Policy answers it. Google Cloud is the exception on that row, because it is the infrastructure Lenz runs on rather than a service Lenz sends text to.
Where the location reads Global, or any Google Cloud region worldwide, Lenz calls that provider’s global endpoint: the provider may run the request in any region it operates in, and gives no data-residency guarantee. Anthropic is the one split case — it stores in the United States but may process in any of the four regions named.
The reasons differ. Anthropic and OpenAI keep what they keep to investigate misuse of their own services, or where the law requires it. Perplexity keeps each request and reply so that a later request can continue from it. Google keeps only the queries sent to Google Search when a check is grounded, to debug its search systems, whether or not anything was flagged. Anthropic, Google, OpenAI, Perplexity and xAI do not train their models on what Lenz sends them.
Data derived from your text
These providers never receive the body of the text you submit. They receive something written from it — a search query, a name to look up — or, when you submit a link instead of text, the link itself. When you check a draft’s citations, they receive the cited links and DOIs, and the page of a statute, regulation or EU instrument the draft cites, not the draft. The timestamping services receive only a hash, which carries no text and no personal data.
| Provider | What it receives | Purpose | Location |
|---|---|---|---|
| Web search providers: Exa, Firecrawl | Search queries derived from the claim, links you submit or that sources point to, the links a checked draft cites, the page of a statute, regulation or EU instrument it cites, and the opinion page of a case it cites | Web search and reading web pages | United States |
| International DOI Foundation (doi.org) | The DOI only; no account data or submitted text | Checking that a DOI cited in a checked draft is registered | United States |
| Crossref | The DOI only; no account data or submitted text | Looking up the registry record of a DOI cited in a checked draft | United States |
| OpenAlex (OurResearch) | The DOI only; no account data or submitted text | Looking up the registry record of a DOI cited in a checked draft | United States |
| CourtListener (Free Law Project) | The case citation only, when Lenz’s own index does not hold it; no account data or submitted text | Checking that a case cited in a checked draft exists, and reading its opinion | United States |
| Wikimedia Foundation (Wikidata) | Names of the people, organisations, places and events a claim mentions | Disambiguating entities that share a name, against the Wikidata catalogue | United States |
| Webshare | Requests to fetch YouTube transcripts; it does not see the video or the transcript | Network proxy; used only when you submit a YouTube link | United States |
| Namirial (qualified trust service provider) | A SHA-256 hash of a warranty certificate: no text and no personal data | Fixing the time a certificate was issued; used only when one is issued | European Union (Italy) |
| OpenTimestamps calendars | The same SHA-256 hash, which they publish permanently | A second, public record of the time a certificate was issued; used only when one is issued | Public servers run by independent operators |
Account data
These providers process account data: your email address, billing details, how you use Lenz, errors our systems hit, and the messages you send us. They do not receive the text you submit for checking.
| Provider | What it receives | Purpose | Location |
|---|---|---|---|
| Stripe | Billing details and payment information | Payments and subscriptions | United States |
| WorkOS | A pseudonymous account identifier and connection details | Authorization when you connect an AI assistant | United States |
| HubSpot | Your email address, plan and usage of Lenz, contact-form messages and chat messages | Customer support and account management | European Union (Germany) |
| SendGrid (Twilio) | Your email address and the emails we send you | Sending sign-in codes and account email | United States |
| Discord | Account events and contact-form messages | Notifications to the Lenz team | United States |
| Sentry | Error reports: the type of error and where in our code it happened, which request it came from, and your account identifiers. Each report is built from a fixed list of fields, none of which carries the text you submit | Error monitoring | European Union (Frankfurt, Germany) |
Error reports Sentry stored before 25 September 2026 can contain the text being processed when the error happened. They are deleted when Sentry's retention period ends.
Changes
We update this page when a provider is added or removed, and when what a provider does with your text changes; the date at the top shows the last change. How we handle your data is described in the Privacy Policy and, for the API, in the Data We Process section of the Terms of Service. Questions go through our contact page.