Privacy and data
What we hold, and what we refuse to
You are researching an employer while you may still work for one. That makes the shape of this page a product decision rather than a compliance exercise.
Last updated 22 August 2026
What is stored
- Your email address and a hashed password, so you can sign in.
- Your answers to the 28 assessment items, and the profile scored from them.
- The reports you have run: the company name and website you typed, the evidence found, and the result.
- A record of what each report cost us to research, in tokens and searches.
What is not stored, and will not be
No resume, and no upload of one. The product does not need it. Not holding a document with your full employment history, address and phone number is a position rather than an oversight, and the day it becomes a feature is the day this page changes.
No tracking of which companies you looked at, for anyone but you. There is no analytics product watching this, no advertising pixel, and no third party that receives your searches.
No employer ever sees anything. There is no employer-facing product, no employer login, and no report about a company that any company can buy, subscribe to, or request. No employer pays us, which is the structural version of this promise rather than the polite version.
Who can read your rows
You, and nobody else with an account. Every table that holds your data has row-level security with owner-only policies, applied in the same migration that creates the table. That means the rule is enforced by the database rather than by the application remembering to filter, so a bug in a query cannot expose somebody else's rows.
It is checked rather than assumed. A script tries to read, insert, update and delete every table with no session, and separately as a second account, and expects every single attempt to be refused. It runs after any change to the schema.
Who else touches your data
Supabase
Database and authenticationStores everything listed above. Your password is hashed by their authentication service and is never visible to us.
Anthropic
The research itselfWhen you run a report, the company name and website you typed are sent to Anthropic's API, along with the instructions for how to research and write it. Your email address is never sent. Your assessment answers are never sent: the scoring that compares you to a company is arithmetic that runs on our side, not a question put to a model.
Vercel
HostingServes the site. Standard server logs, which include IP addresses, are retained by them.
We use Anthropic's paid API, where submitted content is not used to train models. This was one of the reasons for choosing it. A free tier that reads your data to improve a model is not compatible with the paragraph above, whatever it saves.
Deleting everything
One button on your account page. It removes your profile, your answers, every report and every piece of evidence attached to them, and the account itself, in a single transaction. It happens immediately. There is no thirty day window, nothing is held in a backup to be restored later, and there is no support address that can undo it.
You are asked to type the word DELETE first, because that is the right amount of friction in front of something irreversible.
One thing survives, with your name taken off it: the record of what your reports cost us to research, in tokens and searches. It keeps no company name, no result and no link to you. It exists because deleting an account must not make money we already spent look unspent, and because that is an accounting record about our own bill rather than information about you.
Taking it with you
Your account page exports everything the account holds as one JSON file, with no request form and no wait.
Contact
Questions about any of this, including a request to see exactly what is held against your account, go to the address on the account page. This is a small product and the answer will come from a person.