AI policy
No model runs when you use this site, and no model decides what a listing says. Software wrote much of the code that does. All three of those are worth stating plainly, because the index exists to be checked.
Nothing here calls a model
A score is arithmetic over a manifest, by the rules published on the methodology page. Give the same manifest to the same rules and you get the same number, today and next year. The stack check and the report run entirely in your browser and reach nothing at all — the page's content security policy forbids it.
What a model is never allowed to do
- Decide a score, a band or a finding. Those come from the published rules applied to a declared manifest.
- Put a figure on a page. Every number on this site is computed from the crawl or read from a source that is named and dated beside it. A figure never appears because a model produced it.
- Add, change or remove a listing. A listing follows a manifest retrieved from a named origin.
What software did do
Media Yard LLC used AI coding tools to draft much of this site and of Leash Guard, under review. That is disclosed because it is true, and because the alternative — quietly implying it was all hand-written — is the kind of small dishonesty this index exists to find in other people's products. The code is what it is regardless of who typed it: the rules are published, and a listing can be reproduced from the manifest it names.
Why this matters for what we sell
The stack report is a document computed from your own configuration, not a model's opinion of it. If it were the latter it would be worth less and would need a different set of warnings. It is classified, sold and taxed as a digital document, which is what it is. The pre-publish check is computed the same way from the manifests you paste, and the licensed dataset is the index's own records: no model writes, scores or summarises any of them.
What the index does not claim about AI products
Listing a server says what its manifest declares. It says nothing about the model behind it, how that model was trained, or whether
its outputs are safe. Those are outside what a tools/list response can express and outside what this index measures.