Editorial and AI policy
How we use AI to summarise the news, how editors check it, how we credit sources and how we fix mistakes.
Neutral summaries
Each story is one summary of the reporting we found on an event. We stick to what the sources report, attribute claims to whoever made them, and point out where sources disagree. We do not add opinion, and we do not label outlets as left or right.
How we use AI
- Summaries and translations are drafted by a language model that runs on our own hardware. Article text is not sent to third-party AI services.
- The model only works from the articles listed as sources for that story.
- Images marked as illustrations are generated by AI and do not show real events or people.
- Every story notes that it was written with AI assistance.
Human review
An editor reads every story before it is published and checks it against the sources. AI can still make mistakes, so for anything important, follow the source links and read the original reporting.
Sources and credit
Every story lists the articles it is based on, with the outlet's name, its own headline and a link. Some stories rest on a single report; the source list shows that. We summarise and link; we do not republish articles in full. Outlets that want their reporting handled differently can reach us through the contact page.
Corrections
If we get something wrong, we fix it and update the story's date. Report an error through the contact page with the story link.