Newsroom AI Rules Compared: What Nine News Organizations' Published Policies Permit and Prohibit
Newsrooms have published rules for how reporters and editors may use generative AI. This sets out nine of them side by side, from the BBC and WIRED to CBC and a Swiss public broadcaster: whether AI text may be published, what must be disclosed, who signs off, and what is banned, as each organization states it.
As of · Each organization's own policy page, read November 2026; German texts translated by us1
| Organization | Policy and date | May AI-generated content be published? | Disclosure | Human review | Stated limits and bans | Source |
|---|---|---|---|---|---|---|
| BBC (UK) | Guidance: The use of Artificial Intelligence, BBC Editorial Guidelines (undated page) | Not to "directly create news content" for BBC News or factual journalism unless AI is the subject and its use is illustrative; graphics exemptions "may be considered subject to a piloting process" | "must be transparent and clear to the audience" | "active human editorial oversight and approval"; a senior editorial figure is accountable | News video and stills "must not be manipulated beyond a sympathetic crop and minor adjustments"; AI should not be "relied on as a research tool"; AI-generated interview questions to be avoided | 1 |
| WIRED (US) | How WIRED Will Use Generative AI Tools, updated May 22, 2023 | "We do not publish stories with text generated by AI" unless that is the story, and "text edited by AI" is not published either; AI images and video only under conditions | Disclosed for AI images and video | An editor approves any AI-suggested headline or social text | No AI images "instead of stock photography" | 2 |
| The Philadelphia Inquirer (US) | AI Usage Policy, updated December 3, 2025 | Stories with AI support must be "edited and vetted by human journalists" | Required if AI "meaningfully" shapes reporting; "always disclose any known use of AI in illustration" | "every published piece... remains under human editorial control" | "We never use AI to alter news photography or video." | 3 |
| Tagesspiegel (Germany) | Redaktionelle KI-Richtlinien, as of July 17, 2026 (in German) | "We do not publish AI-generated texts with a byline"; AI revision of a text is allowed but does not replace a human edit | Required when AI was "substantially" involved | Final sign-off in the newsroom for original journalistic formats | No AI-generated texts with a byline | 4 |
| WEB.DE / GMX (Germany) | Guideline on dealing with artificial intelligence in the editorial office, updated August 26, 2026 (in German) | AI used for research, audio transcription, headlines, fact-check support and news-curation drafts, "always finally checked and approved" by editors | Required when AI is substantially involved in a text | Editorial approval of curation drafts | No AI images for news unless the images are themselves the subject | 5 |
| swissinfo.ch (Switzerland) | AI guidelines, version 2, updated November 10, 2025 | AI used for transcripts, translations, text editing, summaries, social posts and image, audio and video editing | "Significantly processed" content must be declared; the editor-in-chief decides borderline cases | "checked by a person before publication"; fact-checking mandatory | AI is never the sole research tool; personal data, copyrighted and confidential material only in tools the organization has tested and validated | 6 |
| TaiwanPlus (Taiwan) | Guidelines for Using Generative AI in the TaiwanPlus Newsroom, dated 05/06/2024 (day and month order not stated) | "strictly forbidden from generating reports and stories using generative AI"; used for transcripts and for optimizing social posts and headlines | AI-created images and other content "will be made clear to viewers" | "All AI-generated content is reviewed and edited by our newsroom staff before publishing" | AI treated "the same way we would treat an unverified source" | 7 |
| CBC / Radio-Canada (Canada) | Editor's blog: How CBC News will manage the challenge of AI, June 12, 2023 (preliminary guidance; may be superseded) | AI text or images "only with full disclosure and only with the advance approval of the standards office" | Disclosed before the audience reads, views or hears it | "No CBC journalism will be published or broadcast without direct human involvement and oversight" | No AI recreation of a journalist's voice or likeness (narrow exception); "We will not feed confidential or unpublished content into generative AI tools for any reason" | 8 |
| The Globe and Mail (Canada) | Standards editor's column, The Globe has updated its newsroom AI guidelines, November 1, 2025 (the column says the full guidance is no longer on the website) | "AI can only ever be used with a human in the loop and only in ways that assist journalists, not to replace them" | AI contributions "must be clearly labelled, with a description of how AI has been used" | Human in the loop | Generative AI is "unsuitable for core writing and editing work" and "not a reliable research tool"; unpublished stories cannot go through AI tools outside cases cleared by the legal team | 9 |
Each cell quotes or closely paraphrases the organization's own page; quotation marks are the organization's words (translated for the three German-language pages). Policies change, and some pages are undated, so the dates are those shown on the page. This list covers rules, not practice: it does not report whether any organization has followed them. Which organizations appear depends on which policy pages we could open; the Associated Press, the Guardian, Reuters, the New York Times and the Financial Times are absent because their pages were blocked or not located, not because they have none. Rules are not graded by our evidence grades.
The analysis
Highlights
Text is where the rules differ most
WIRED says it does not publish stories with AI-generated text, or text edited by AI, unless that is the story;2 TaiwanPlus is "strictly forbidden" from generating reports and stories with generative AI;7 Tagesspiegel publishes no AI-generated texts with a byline but allows AI revision that does not replace a human edit;4 CBC allows AI text only with full disclosure and the standards office’s advance approval.8
The BBC bans by purpose, not by tool
Its guidance says generative AI should not be used to directly create news content for BBC News or factual journalism unless AI is the subject and its use is illustrative, allows tools that "provide information, insight or analysis" to aid production, and warns they should not be relied on as a research tool.1
Disclosure thresholds vary
The Philadelphia Inquirer requires disclosure if AI "meaningfully" shapes reporting; Tagesspiegel and WEB.DE when AI is substantially involved; swissinfo for "significantly processed" content, with spell-check and translation adaptation exempt; CBC before the audience reads or hears the content.34568
Confidential material gets its own rule in only some policies
CBC says it will not feed confidential or unpublished content into generative AI tools for any reason; swissinfo allows such material only in tools it has tested and validated; the Globe and Mail says unpublished stories cannot go through AI tools outside cases cleared by its legal team.689
Photos and video are treated more strictly than text
The Inquirer says it never uses AI to alter news photography or video; the BBC says news video and stills must not be manipulated beyond a sympathetic crop and minor adjustments; WIRED does not use AI images instead of stock photography.123
Why this matters for the reporter’s job
News analysts, reporters and journalists are SOC 27-3023.00 in O*NET, which counts 47,100 of them in the US in 2025 with a projected decline of 1% or more to 2035 (BLS projection, not an estimate of AI’s effect).10 These policies show which of the role’s tasks newsrooms fence off from AI (writing and publishing), which they open to it under human oversight (transcription, headlines, translation, summaries) and where they ask for care (research). For how AI is graded across other roles, see how we measure AI in roles.
Sources
- BBC, Guidance: The use of Artificial Intelligence, BBC Editorial Guidelines (undated).
- WIRED, How WIRED Will Use Generative AI Tools, updated May 22, 2023.
- The Philadelphia Inquirer, AI Usage Policy, updated December 3, 2025.
- Tagesspiegel, Redaktionelle KI-Richtlinien: So arbeiten wir mit Künstlicher Intelligenz, as of July 17, 2026.
- WEB.DE (1&1 Mail & Media), Leitlinie zum Umgang mit Künstlicher Intelligenz in der Redaktion, updated August 26, 2026.
- SWI swissinfo.ch, AI guidelines, version 2, November 10, 2025.
- TaiwanPlus, Guidelines for Using Generative AI in the TaiwanPlus Newsroom, dated 05/06/2024.
- CBC News, How CBC News will manage the challenge of AI, Editor’s Blog, June 12, 2023 (preliminary guidance).
- The Globe and Mail, The Globe has updated its newsroom AI guidelines, standards editor’s column, November 1, 2025.
- O*NET OnLine, 27-3023.00 – News Analysts, Reporters, and Journalists (employment and projection from the Bureau of Labor Statistics), retrieved 2026-11-09.