Privacy
Last updated: 2026-11-06 (contact is by email only; list of services updated to match current operation). API retention statements unchanged since 2026-10-16.
What we collect
The content you send to our API, and basic request information (time, page, browser type, and your IP address, shortened before we store it). Log lines for unreadable model output carry up to 200 characters of that output. Fingerprints keep verdicts, providers, observation counts and timestamps, no text.
How long we keep it
The /evaluate result cache is disabled: new result-cache entries are not written and existing entries are not served. Previously written entries may remain until their existing expiry, up to 24 hours after writing. Claim fingerprints retain hashes, verdicts, providers, observation counts and timestamps without an automatic expiry; they do not contain claim text. Checker logs retain claim hashes and checker verdicts for up to 180 days. Separate application and hosting logs may retain excerpts of submitted text or model output. Request records: 30 days. Accuracy statistics, which contain no text: up to 180 days.
Anchoring
The SHA-256 hash of each queued receipt is published as part of a Merkle root on GOAT Network, a public blockchain; no claim text or receipt content goes on-chain, and on-chain data cannot be deleted.
Who helps us run it
Vercel and Upstash (hosting and database, US), Perplexity and OpenRouter (the model-based assessments behind /evaluate; rule-based checks use neither), Google Workspace (email), Google Fonts and Fontshare (website fonts), and GOAT Network (public anchoring of receipt hashes, below).
No longer used. No payment provider or API gateway is in use today, and no API route takes payment. Until 24 September 2026, a Zuplo API gateway and a Stripe-billed self-serve tier were in use; both were withdrawn that day. Until 4 October 2026, paid research routes accepted x402 payments through payment facilitators; those routes were retired that day.
What we don't do
We don't sell your data, we don't train AI models on it, and our website uses no analytics or tracking cookies.
Emailing us
There is no inquiry form. The contact page shows our email address and a link that opens your own email application. Nothing you write is sent through this website.
What we receive. Whatever you include in your email, along with your email address and the usual email headers. It arrives in a Google Workspace mailbox (joe@tanilo.io).
Why. To reply to you about whether Tanilo is a fit. We don't add you to a mailing list.
How long we keep it. Emails are kept in the mailbox as correspondence. We have not set an automatic deletion period.
Deleting it. Email joe@tanilo.io from the address you used to ask for a copy of your messages or for them to be deleted from the mailbox.
Please don’t include confidential documents, passwords or sensitive personal information. We agree how to handle samples separately, before you share any.
Your choices
Email joe@tanilo.io to ask us to remove a claim you sent, get a copy of what we hold, or request our detailed data-handling documentation.
This is a plain summary, not a formal privacy policy.