Privacy

What this site is

A green-claims scanner for online shops, free to use and hosted at cosmeticlaim.eu. It takes no payments and has no user accounts.

Who is responsible

This site is operated by Distiller OÜ (registry code 14319231), Piima tn 1, 11317 Tallinn, Eesti — Jimmy Karp is the person behind it. In data-protection terms the company is the controller: it decides what happens to everything described on this page. tere@cosmeticlaim.eu reaches him directly; there is no support desk in between.

What happens when you scan something

• A URL you submit is fetched by our crawler, the text is analysed in memory, and the result is returned to your browser. The pages are not stored.
• Text you paste is sent to the scanner, analysed, and discarded when the response is sent. It is not written to disk and not logged.
• An image you upload is held in memory for the length of the request while text is extracted from it, then discarded. The file is never written to disk.
• The report exists only in your browser. Reload the page and it is gone.

Where your text goes

This is the part worth reading twice. Detection runs on our own server, but the second stage — deciding which findings are marketing rather than commentary, and drafting the rewrites — runs on a language model hosted at ai-gw.nurme.eu, a private gateway on infrastructure operated by the same person as this service.

• What is sent: for each finding, the matched sentence (up to 320 characters), the term that matched, the path of the page it came from, the language the text is written in, and our own note on why that term is restricted — the same note the report shows you. Nothing else from your text: a sentence that matched nothing never leaves this server, and neither does the query string of the URL.
• What is not sent: the rest of the page, the full text you pasted, your image, your IP address, or anything identifying you.
• What is kept: nothing by us. The gateway keeps operational logs of request volume and token counts, not request bodies.

If you would rather no part of your text left this server, that is a reasonable position. Say so and a deterministic-only mode is a small change — the rule engine works on its own and the report already shows what it found before the model saw it.

What is logged

The scanner keeps standard operational logs: timestamp, request path, response status and duration, and — for failed scans — the URL that failed and the reason. Your IP address is held in memory for rate limiting, in a counter that is cleared roughly every hour. Pasted text and uploaded images never appear in logs. The server also writes one line of aggregate counters per hour — how many requests, how many distinct visitors, which verdicts, which interface languages — so we can tell whether anyone uses this at all. That line holds no addresses, no URLs and no text: visitors are counted through a hash that is salted per hour and discarded with the hour, so the number survives and the identifier does not.

How long anything is kept

Your text, URL or image: until the response is sent, then gone. The rate-limit counter: in memory only, entries older than an hour are dropped, and nothing is written to disk. The server journal: method, path, host, status and duration — it names no person, and it rotates by size rather than on a schedule. The hourly totals file: kept indefinitely, because it holds no addresses, no URLs and no text.

Why we are allowed to do this

Two bases, one per operation. What you submit — text, URL or image — is processed under Article 6(1)(b) of the GDPR: it is objectively necessary to produce the report you asked for, and there is no report without it. The technical minimum around that, the in-memory rate-limit counter and the hourly totals, runs under Article 6(1)(f), legitimate interests — keeping a free service available against abuse, which Recital 49 names as network and information security, and knowing whether anyone uses it at all. Neither interest needs to know who you are: the counter never reaches disk and the totals hold no identifiers. When the crawler reads a shop you pointed it at, personal data on those public pages reaches us from the page and not from the person, and writing to each of them would be disproportionate — Article 14(5)(b). There is no profiling and no automated decision with legal effect for you.

Cookies and tracking

None. No cookies are set, no analytics script runs, no third-party requests are made from the page. Fonts and styles are served from this domain. The usage counters described above are tallied on the server and need nothing stored in your browser — they cannot follow you between sessions or across sites.

Sites we visit on your behalf

When you scan a URL, our crawler appears in that site's logs as CosmeticlaimBot/1.0, identifying this project and linking to the page describing it. It reads only public pages, follows robots.txt, and stays within the limits documented there. CosmeticlaimBot

Your rights

Since nothing you submit is retained, there is generally nothing to access, correct or delete. If you believe something about this site's handling of data needs attention, see the contact section of the terms. Terms of use

If you think this site handles data wrongly, write to us first — but you are not required to. You can complain directly to the data protection authority of the EU country where you live, where you work, or where you believe the problem happened.

Changes

If this build ever starts retaining scans — for opt-in benchmarking, for example — this page will say so explicitly before that happens, not afterwards.