The switch is on when you sign up. Turning it off is starting to cost money.
Companies have mostly stopped announcing that they train AI models on your data. The pattern now is quieter: the setting is already on when you sign up, and whether you're allowed to turn it off depends on which plan you're paying for. This site tracks that pattern, company by company, in the companies' own words.
Who runs this
One person, Josh, based in the UK. I can't lie and say that the entirety of the website is hand-written. A language model drafts most of the words on it - the verdict notes, the summaries, a good part of what you're reading now - working from the archived policies of each company. What it can't do is decide. I decide the grades. I check every quote against the archived page before it's published, and the build refuses any quote that isn't there word for word. I write to the companies, I read what comes back and I edit what the model wrote when it's wrong or when it doesn't sound like me. So if you notice the language changing across the site, that's why.
There's no investor and no partner. The only money the site takes is the labelled paid notices, at one price that doesn't change with a sponsor's grade. Press enquiries go to the same address as everything else, contact@isittrainingonme.com, and the press page has the figures, the boilerplate and images you can lift.
What we promise about every claim
Every verdict quotes the company's own policy word for word, never paraphrased, with a date, the plan it applies to, the jurisdiction it applies to, and a link to an archived copy anyone can check. If a build would publish a clause we can't verify word for word against that archive, the build fails. See methodology for the full rubric.
How it's built
A separate monitor fetches each registered policy page on a schedule, archives what it finds, and diffs it against the previous version. That's the change feed. This site is a renderer over that archive. It doesn't re-fetch anything itself, and every clause it displays traces back to a specific, dated snapshot on file. See /bot if our traffic showed up in your logs.
What we don't do
- — No user accounts, and no automated opt-out performed on your behalf.
- — No inference about what a policy "really means." UNCLEAR means the text doesn't say, and we say so.
- — No grade changes because a company complains. A grade changes when the quoted clause changes — see /advertise for how sponsorship is firewalled from editorial.
Something wrong, or something to add?
The repository behind this site is private, so the submission form is how you contribute: suggest a service to track, or report a clause that looks wrong. Everything else: contact@isittrainingonme.com. What a grade is, how a company gets a right of reply, and what you may do with the site's work are on the terms page.