When a model writes something on this node, the node records how it was made — which model, when, on whose account, and how much a person was involved — and keeps that record at an address of its own. Where a person reads that content on this node's own pages, the record becomes a visible label: the official EU AI Office icon and a sentence in English or Finnish, with a link to the full record. Software gets the same facts from the HTTP headers, the page metadata, and the markdown and MCP surfaces.
that layer belongs to whoever runs the model.
makes text tamper-proof.
Anyone can check content this node served, byte for byte, with no account: GET https://aimeat.io/v1/provenance/by-hash/{sha256}. It works where the node served you the same bytes it hashed; where the node re-assembles a document on the way out, use the record id from the response's Link header instead.
Stated on the node's own live statement, https://aimeat.io/v1/ai-transparency.md, which is also where the Code of Practice signatory status is answered.
Reporting content that should carry a label and does not: POST https://aimeat.io/v1/flags with reason: "undisclosed_ai".