Is this a way to detect AI? No — the opposite. It's a self-declaration. There is no detector, no account, and no registry. Detection accuses the innocent; a declaration credits the honest.
Who owns it? Can I use it commercially? Nobody owns it. The scale is CC0 (public domain); the site text is CC BY. Use it, quote it, build on it, fork it. There is no committee to ask and no fee.
Does a badge prove a work's level? No. A badge is a public self-declaration by whoever is accountable for the work — like a byline or a nutrition label. It is cheap to make and expensive to break, not a certificate of proof.
Does declaring Level 4 make me EU AI Act compliant? No. The scale is not a compliance product and this is not legal advice. The law's reviewed/ unreviewed line happens to match the Level 4 / Level 5 distinction, but whether a duty applies to you is a question for a lawyer. See /eu-ai-act.
How is this different from C2PA / Content Credentials? Complementary. C2PA gives cryptographic, tamper-evident provenance about an asset — what touched this. The scale is a declaration about contribution and review — whose thinking is inside this, and who stands behind it — and it works for plain web text, not just media files. Every level also maps to an IPTC Digital Source Type, so a declaration can travel into C2PA/XMP metadata.
Is it a W3C / IPTC / C2PA standard? No. It interoperates with their work and requests their review; it never claims their endorsement. It's an independent, open standard offered as input to that conversation.
Why does it start at zero? So human-made work has a number too. A system that labels only AI makes everything unlabelled read as human by default — including all the AI it missed. Level 0 is load-bearing, not a courtesy.
Why can't a "higher" level be better? Because the moment one level becomes an insult, everyone lies their way down to a "safer" number and you've rebuilt the broken binary with extra steps. No level ranks above another. Level 5 is the honest answer for an automated market report; Level 0 is the honest answer for a memoir.
What data do you collect? None. No analytics, no cookies, no fonts or badges fetched from us (badges are inline SVG, so they can't phone home). Signing is a pull request — there's no email list to leak. See docs/PRIVACY.md.
What happens when a work won't classify cleanly? That is a finding about the scale, and we want it. Content and classification questions are public GitHub issues; the case corpus exists precisely to collect the hard ones.
Can it be misused? Yes, like any declaration — and docs/THREAT-MODEL.md is honest about how, what the design mitigates, and what it deliberately does not try to solve.