Privacy Policy

Last updated 22 September 2026

This policy explains what the Likely website, API and add-ons for Google Sheets and Microsoft Excel ("Likely") collect, why, and what we do with it. Likely is operated by Simber Tech ("we"), based in the Netherlands. Contact: support@simbertech.com.

The short version

What we collect

Account data

When you request a key, we store your email address, the key, your plan, the date and, if you arrived from a link we tagged, the campaign name, and we email the key to that address. You can ask us to send it again from the support page. If you subscribe, Stripe processes your payment and we store your Stripe customer and subscription IDs. We never see or store your card details.

Usage data

We count the rows judged per key per month to apply plan limits.

Spreadsheet content

When you use a formula, bulk run, the live demo or the API, the cell text and your question are sent to our server and on to TypeSafe, our AI provider, to be judged. We do not store the text of your cells. To make repeated cells free, we store a one-way cryptographic fingerprint (SHA-256) of your question, the labels or levels you gave and the cell text, alongside the answer (for example, a probability), for 30 days from the day that answer was worked out. The fingerprint is not reversible, but someone who already held both the fingerprint and a good guess at the text could confirm that guess, so we treat fingerprints as data derived from yours rather than as anonymous.

The same question and the same text always give the same fingerprint, and that store is scoped to your key: an answer worked out for your account is reused only for your account. The live demo on our website has its own separate store, shared between the anonymous visitors who use it and holding nothing that belongs to an account. No account can read another account's cells, answers, keys, usage or bulk runs.

Bulk run results are stored for 30 days so the add-on can fetch them. They hold the answers and their row order, under your key, so we can write them back to the right rows. They do not hold the text of the cells you judged: where a formula returns one of your own labels, each answer records which label it was rather than the label itself. A bulk run does keep the list of labels you gave beside the job for those same 30 days, so the answers can be written back under the right names; nothing keeps them after that, and the cache never holds them at all. Both clear-outs are done by a sweep that runs on our server once a day.

Add-on permissions

The Google Sheets add-on only accesses the spreadsheet it is opened in (spreadsheets.currentonly), shows its sidebar (script.container.ui), and connects to Likely's server (script.external_request). It stores your key in that spreadsheet's document properties. The Excel add-in stores your key in the add-in's local storage on your device. Neither reads other files, email or contacts.

Website

The website sets no cookies at all. We measure visits with PostHog in cookieless mode: it stores nothing in your browser and counts visitors with a rotating one-way hash calculated on PostHog's servers in the EU, so there is no profile to link back to you and nothing to consent to. We also count a few product moments this way, such as a demo run or a key request. The site does store your key in your browser's local storage after you request one, so the upgrade buttons work. If you arrived from a link we tagged with a campaign name, the site also keeps that name for the length of the visit and sends it with a key request, so we can tell which post or listing people found us through. It is a word like “hn” and it says nothing about you. Your browser forgets it when you close the tab; if you request a key during that visit, we keep it on the key's record, as listed under Account data above. Our server keeps standard request logs (IP address, time, path) for security and rate limiting.

Who processes data for us

TypeSafe AI, Inc. (United States)Our AI model provider, and the only one. Our server calls their System One API directly (api.typesafe.ai, model jev-1.13.0) on our own account, with no aggregator, gateway or model hub in between. They receive your question, the labels or levels you gave and the text of the cells you run, and return an answer. Their privacy policy states they will not train or fine-tune any AI or machine learning model on the prompts we send, and we have not given, and will not give, permission for any such use.
StripePayments and subscription management.
PostHog (EU)Cookieless website and product analytics. Receives page and event data, never your cell text.
Hetzner Online (Germany)Hosts our server and database.
CloudflareDNS and proxy in front of the website and the API. Because the proxy ends the HTTPS connection, cell text passes through it on its way to us. We use no Cloudflare AI, gateway or worker product: it carries the traffic and nothing more.
Namecheap Private EmailDelivers your key when you sign up or ask for it again. Never receives your cell text.
Google CloudReceives error reports from the Google Sheets add-on, through Apps Script's own logging. These are error messages, never your cell text.
Google FontsServes the website's fonts.

These are the providers we use today. If we change one we update this page before the change takes effect, and we will only ever use an AI provider whose terms commit not to train on what we send, and from whom we withhold any consent to such use.

Use of Google user data

Likely's use and transfer of information received from Google APIs adheres to the Google API Services User Data Policy and the Google Workspace API User Data and Developer Policy, including the Limited Use requirements. Spreadsheet content is used only to provide the judging feature you request, is not used for advertising, is not sold, and is not read by humans unless you ask us to help with a support request and share it with us.

We do not use, transfer or sell Google user data — raw, aggregated, anonymised or derived — to develop, improve or train any non-personalised, foundational or generalised machine learning or artificial intelligence model. This applies to us and to everyone we send the data to. We build no model of our own, personalised or otherwise, and we do not fine-tune, adapt or personalise anyone else's.

Cell text goes to exactly one third party: TypeSafe, named above. We send it so an answer can come back to you, and for nothing else. Their published privacy policy states that prompts and other input are not used to train or fine-tune any AI or machine learning model, and we have withheld consent to any such use. We call their API directly from our own server with our own key, and on our side there is no aggregator, gateway, router or model hub in the path. We run no AI model of our own: no model runs on our server or on your device.

How long we keep data

Account and usage data: while your key exists, then deleted within 30 days of your request. Cached fingerprints and answers: 30 days from when the answer was worked out. Bulk run results: 30 days. Both are cleared by a sweep that runs on our server once a day. Server logs: kept on our own server and never sent anywhere else. They record request metadata such as IP address, time and path, never your cell text, and we delete them once they are no longer needed for security and rate limiting.

Your rights

You can ask to access, correct, export or delete your data by emailing support@simbertech.com. If you are in the EU, UK or a similar jurisdiction, you also have the right to object and to complain to your data protection authority. Our legal bases are performing our contract with you (providing Likely) and legitimate interests (security, abuse prevention).

Security

All traffic uses HTTPS. API keys for our providers never leave our server. Access to production data is limited to the operator.

Changes

We'll update the date above when this policy changes, and email subscribers about material changes.

Simber Tech · the Netherlands · KvK 94426392 · VAT NL005083111B87
simbertech.com · support@simbertech.com