Vice Atlas

1. Sourcing standards

Every factual claim on Vice Atlas is sourced. We treat sources in a three-tier hierarchy:

Tier 1 — confirmed. Stated by Rockstar Games, Take-Two Interactive, or visible on-screen in an official Rockstar trailer or Newswire post. Only Tier 1 information is ever labeled confirmed in our writing.
Tier 2 — leaked. Material from a documented leak (e.g., the September 2022 GTA VI development footage). We cover leaks editorially, but we never republish leaked copyrighted footage and we always label leaked material as such — never as confirmed.
Tier 3 — rumor / fan attribution. Casting speculation, voice-actor attributions sourced to LinkedIn or social-media patterns, fan-decoded easter eggs. We will discuss these but never as fact, and always with the explicit note that the source is unverified.

2. Speculation vs. reporting

When a story warrants editorial speculation — about Rockstar's marketing rhythm, Take-Two's likely launch sequencing, or what a particular trailer beat might foreshadow — speculation is clearly framed in its own section, separated from the confirmed-facts section, and written in our own voice rather than attributed to anonymous "sources said."

We do not run unsourced "sources tell us" reporting. If we cannot name a source or point to a verifiable artifact, we do not publish the claim as fact.

3. Leak handling

Leaks are part of the GTA VI news cycle. We cover them editorially — analyzing what they imply, what they confirm, what they don't — but we operate within these limits:

No leaked footage embeds. We do not embed, host, or link to leaked development footage. Citations to leak archives are limited to clearly-attributed text descriptions.
No leaked source code or assets. Same standard.
No attribution that endangers leakers. We do not name individuals connected to leaks beyond what is already public in mainstream press.
Removal on takedown. If Rockstar Games or Take-Two issues a DMCA takedown for any specific leak-derived content, we comply and document the removal in a public corrections log.

4. AI usage

We use AI tools for research, outlining, fact-cross-referencing, and copy editing. We do not publish AI-generated articles as original analysis. Specifically:

Allowed: AI-assisted research summaries (verified against primary sources before use), outline generation (rewritten in human voice before publishing), grammar and style checks, alt-text generation, schema markup generation.
Not allowed: AI-generated final copy passed off as human-written; AI-generated factual claims unverified against primary sources; AI-generated images presented as original photography or screenshots; AI-generated "quotes" attributed to real people.

Every published article is written by a human author who is named on the byline and accountable for every word, regardless of which tools assisted the drafting process.

5. Attribution and plagiarism

We credit every source we materially relied on, by name and link, in the article body or in a Sources section. If another publication broke a story and we are following with our own analysis, we link to and credit them — even when we add new reporting or analysis on top.

We do not republish other publications' original reporting without explicit attribution and a link to the original. We do not paraphrase other writers' analysis closely enough that an attentive reader could reasonably mistake it for our own.

6. Conflicts of interest and disclosure

We have no advertising relationships with Rockstar Games, Take-Two Interactive, any GTA-adjacent publisher, or any party with a material interest in our editorial coverage. We accept no review embargoes, no exclusive access in exchange for favorable coverage, and no sponsored editorial.

If any of these change, the relationship will be disclosed at the top of every relevant article and on the relevant directory pages.

Affiliate links to gaming peripherals (controllers, headsets, chairs, monitors) are clearly labeled inline using a standardized disclosure phrase. Affiliate links never appear in news coverage — only in dedicated buying guides where the relationship is the point of the article.

7. Corrections

When we make a factual error, we correct it publicly and we document the correction. We do not silently edit published claims. Full process documented in the corrections policy.

8. The board's integrity rules — locks, voids, and grades

Every call on the board carries its resolution rule, source, and lock date, written before anyone taps. These rules govern what can change afterward — published here first, so any future change is judged against a policy that predates it:

Locks obey reality. Every call locks at its stated time or the instant any official information bearing on its outcome publishes — whichever comes first. A call is never tappable after material information lands.
Lock changes are narrow and public. A lock date may be extended only before the original lock passes, only while fewer than 10 positions are recorded, and never after any material official event. Every change writes a permanent, dated note on the call page. Locks are never shortened, and a call that has locked is never reopened.
Voids are for broken questions, never losing sides. A call is voided only when reality answered it before its lock (a call nobody can lose is not a call), its resolution source ceased to exist, or its rule proved unadjudicable. Voided calls stay visible forever with their written reason.
Grades are immutable. A graded call is never re-graded; a grading error is corrected as a new, dated note — never an edit.
We have no sources. Vice Atlas never solicits, pays for, receives, or publishes non-public information. Calls about Rockstar's calendar print their reasoning from public precedent on the card. Calls on whether leaks, hacks, or street-date breaks occur are forbidden, permanently.
The Insider Index grades claims, never people. The Index records public claims by named sources with large audiences — quoted verbatim, dated, archived at entry — and grades them against official outcomes only. Hedges are recorded as hedges, never scored. Any graded source's on-record response is published with their entry, and factual errors are corrected within 24 hours.
The desk is machine-assisted, human-graded. Scoreboard rows and standard cards are assembled by our own tooling from the database and official sources; a named human reviews and holds editorial responsibility for everything published. Nothing generated is presented as a human byline, and no generative imagery is used anywhere, ever.

9. Updates to this policy

This page is the canonical statement of how Vice Atlas operates editorially. When the policy changes, the change is documented with a dated note on this page rather than rewritten in place.

2026-07-29: added section 8 — the board's integrity rules (locks, voids, grades, the Insider Index, the desk).

Last reviewed: 2026-07-29. Questions about this policy: email hello@viceatlas.com.