About and methodology

DRAFT for the invited beta. The sections in [brackets] are still to be written.

Who is behind it

[To complete: Sushil: who you are, why you're building it, a photo, how to reach you.]

How a result is made

1. An AI reads the label and returns the words printed on it. It never decides anything.
2. A second, independent reader checks the photo for additive numbers the first one may have missed. It can only add a number printed on the pack, never remove one.
3. Fixed rules check each ingredient against our tables. Every additive in the tables is checked against the FSSAI regulations and the Codex international list, with the date it was checked.
4. When the photo is too hard to read, the app says so and asks for a clearer one, rather than guessing.
5. A person reviews every new pack against its photos before it joins the catalogue. After that, scanning its barcode shows the reviewed result, with the date it was last checked.

Rules in other countries are shown only as extra information, when you ask for them. They never change a result.

Money and brands

A brand can never pay to change, soften or preview what InsideLabel shows about its products. [To complete: the full conflict-of-interest policy, before any revenue.]