Responsible AI Usage Guide

AI tools can help a newsroom work faster, but they can also produce confident errors. This guide describes how the BreakingNews24hr demonstration newsroom intends to use them. Stories on this site are fictional and illustrative.

What AI tools may be used for

• Transcribing interviews and recordings, with the transcript checked against the audio.
• Summarising long documents as a starting point for a journalist's own reading.
• Suggesting headline options, spelling and style corrections.

What AI tools may not do

• Serve as a source of facts or quotations.
• Publish anything without human review.
• Generate images presented as photographs of real events.
• Be fed confidential source material through unapproved services.

Human sign-off

A named editor is accountable for every published word, whatever tools helped produce it. Any claim that originated in AI output must be verified against independent sources, following the same audit trail as any other claim on the Fact Pledge page.

Labelling

If AI materially shaped a story, a graphic or an image, readers should be told in plain words. Illustrations that are generated or conceptual must be labelled as such, and never passed off as documentation.

The tool can draft. Only a person can be responsible.

Review and feedback

This guide is expected to change as the technology does. Questions and concerns are welcome at [email protected].