Accept that AI is coming to newsrooms, let’s establish rules for disclosure

Accept that AI is coming to newsrooms, let’s establish rules for disclosure

To opine about the danger of artificial intelligence replacing a local reporter or editor is to re-tread old ground. I accept that AI will become an accepted toolset one day soon and choose instead to focus on the policies we’ll need to adopt to use them responsibly.

Principal among these rules, I think, should be disclosure of whether an AI tool was used in the creation of content.

Done correctly, this labeling can become a centerpiece of transparency alongside a responsible publication’s bylines, datelines and flags indicating content is sponsored or opinion-based. It may also help create trust with an audience already primed to believe that generative AI can be prone to biases and errors.

Furthermore, we should identify the kinds of AI tools used in the work. This could go a long way toward transparency and could help to overcome the decades of Hollywood stereotypes linked to AI.

E.g. Was the illustration generated by an image model based on the author’s prompts? Was the text drafted by a language model? Was the editing or creation of video or audio aided by an AI tool? Did automation provide the transcript of the interview? Was content automation used to reformat content for a certain platform?

In each example, a very simple disclosure would suffice.

Man and robot in a field of flowers, in the style of Rembrandt, generated by DALL-E.

In addition, requiring journalists to track and disclose their AI usage can promote responsible use. A real-life person will be accountable for the final product and therefore inclined to be more careful and deliberate in the use of AI augmentation.

Here are Phil’s three laws* of AI-assisted reporting, in the style of Isaac Asimov’s Laws of Robotics.

  1. Journalists must oversee and disclose the use of AI tools, ensuring the published results are consistent with the second and third laws.
  2. AI tools can augment but must not replace human authors, editors, photographers or illustrators or through their use be allowed to undermine the public respect for those roles.
  3. Use of AI tools should only be allowed when they can meaningfully benefit the audience.

*I admit, I tried using ChatGPT to generate these laws, but wasn’t happy with the results.