Editorial and AI Policy

We collect from configured sources such as official company and research feeds, established reporting outlets, technical publications, public research repositories, and selected community sources. Source configuration and credibility rules live in application code rather than model prompts.

We prefer direct and attributable reporting. A source appearing in the product does not mean we endorse every claim it publishes. Readers should open the original article for full context and updates.

Collected items are normalized into a common story format. We compare canonical URLs, titles, entities, topics, and coverage clusters to reduce duplicates and connect reporting about the same event.

Grouping can be imperfect. Separate events may be combined, and related coverage may be missed. We preserve links to original reporting so readers can compare sources directly.

AI services may assist with summaries, headline or description rewriting, semantic embeddings, topic and entity enrichment, and illustrative image generation. These tools help normalize and organize public news content; they do not replace the linked source.

Automated output can introduce errors, omit nuance, or become outdated. Application checks constrain format and source fields, but no automated safeguard can guarantee accuracy. For consequential decisions, rely on the original reporting and primary documents.

A story image may come from available source media or may be an illustrative asset created for the normalized story. An illustrative image represents the topic and should not be treated as a photograph of the reported event, person, product, or evidence unless its source makes that clear.

The shared edition uses deterministic signals such as recency, source credibility, novelty, trend strength, topic coverage, entity coverage, and duplication controls. These signals are editorial choices encoded in software and may change as coverage quality improves.

Trending, latest, briefing, and directory views serve different reader goals. A high position means the item scored strongly for that view at that time, not that the claim is true or universally important.

For You ranking can use selected topics, followed sources or entities, saved stories, reading history, search memory, explicit feedback, recency preferences, and diversity controls. We aim to show a clear reason for personalized recommendations and give readers controls to inspect, change, remove, or reset signals.

Personalization should not trap readers in one narrow view. Diversity, freshness, source, entity, category, and angle rules help preserve a useful shared news core alongside individual interests.

We record reader signals only from explicit, meaningful actions such as saving, opening, following, searching, sharing, giving feedback, or changing a preference. We do not treat rendering, refreshes, visibility churn, or polling loops as evidence of reader interest.

Each normalized story identifies and links to its source when that information is available. Publishers, authors, photographers, and other rights holders retain their rights in original material. Our summaries and organization are not a substitute for supporting the publishers whose work informs the edition.

We may update, regroup, relabel, or remove normalized material when a source changes, a duplicate is discovered, an automated summary is misleading, or a rights or safety issue is substantiated.

To request a correction, include the page URL, the specific statement or image at issue, the supporting source, and how we can follow up. We review good-faith requests and prioritize errors that could materially mislead readers.

The current product has no paid placement, sponsored ranking, advertising, pricing, or subscription entitlement. If sponsored material or commercial ranking influence is introduced, it will be labeled clearly and this policy will be updated.

No source list or ranking system is complete. Coverage can overrepresent publishers that offer reliable feeds, publish frequently, or use formats that are easier to normalize. We continue to review source diversity, missing context, duplicates, and systematic ranking effects.