First News AI Guidelines

By steve.smith • 24th April 2024

  1. AI can help – but it doesn’t do your job for you. Use it to speed things up, reword something, suggest a headline or to brainstorm, but never let it publish content unchecked.
  2. You’re our writer, not AI. Any content that AI, or even FioNa creates must always be reviewed, edited and approved by a human. As a publisher, we are legally responsible for any AI content.
  3. Accuracy is paramount. All AIs acknowledge they hallucinate (make stuff up). First News is a trusted source of information. Just because ChatGPT or whatever sounds confident doesn’t mean it’s right. Fact-check everything, especially names, dates and quotes. 
  4. If AI was involved, consider labeling it. If a story, headline or image was generated with AI, we should be transparent – it builds trust. Do we need a catch-all disclaimer somewhere? Are images different to text? Does generative fill in Photoshop/InDesign need labelling?
  5. Keep to First News tone. AI can sound too grown-up, robotic or wrong, and sometimes you can just spot it a mile off. Make sure the article you end up with fits with how we talk to our readers. FioNa is programmed to write in First News style, tone and level of comprehension.
  6. Check for weirdness. AI can repeat all sorts of nonsense from the internet. Keep an eye out for bias, especially if you have fed AI a external source.
  7. Avoid AI for sensitive or upsetting stories. If a story needs care and nuance, it needs a human. As a backstop, FioNa is programmed to include our default line for sensitive stories, eg, “Talk to a trusted adult, or contact Childline.”
  8. Keep kids’, and everyone’s, data safe. Don’t feed private info into AI tools. Some data gets used to train AI models. We have opted out of FioNa from sharing data with OpenAI. Google NotebookLM doesn’t share data in any case.
  9. Want to use a new tool for the paper? Drop it in the Teams AI channel or DM/email Steve, Eddy or Nicky, and let’s talk, see what it does. Alternatively do something surprising with it!
  10. We’re all learning. These guidelines will evolve as the technology does. If you have any questions, concerns or suggestions, just drop a line in the Teams AI channel or DM/email Eddy, Nicky or Steve

PS. Have a read of the BBC AI Guidelines. they’re a bit more formal and comprehensive. Do we want our guidelines to be more like theirs or simpler as above? Consider that our audience is probably similar to the BBC’s. What do you think?