279: Things to consider in your AI policy

Some publishers see generative AI as a fundamental challenge to their business. Some would say it’s an “existential threat” but that phrase is overused and annoying. 

Generative AI is certainly disruptive, but it also presents a lot of new opportunities. 

As I say in my most recent letter, I believe the larger threat is from AI-curated content. 

No matter where you stand on this, you should have a written policy for the use of AI in your organization. 

That policy should cover general principles like … 

Transparency – when and to what extent do you have to disclose that you’ve used AI? 

Accountability – to what extent is a human responsible for the final product. 

Data privacy and compliance – make sure you’re not feeding AI with information it shouldn’t oughta have. 

Your policy should also cover how AI can be used. For example … 

• As an assistant in research and analysis. 

• To check grammar and spelling. 

• To create or edit first drafts. 

• To create or enhance images (with some limitations below). 

• To create a transcript of audio or video content. 

In all these cases you should give some guidelines on where AI work ends and a human’s review begins. 

You should list prohibited activities, like … 

• Fully automated content generation. Personally I think fully automated content is fine in certain cases – such as creating summaries of an article – so long as that’s clearly explained. 

• Using AI images to depict real people or events. 

• Using AI to make value judgments on sensitive issues. 

The policy should say whether all content must be reviewed by a human author. That’s particularly important with fact-checking. 

You might also require a certain level of training before employees are allowed to use AI. 

I’ve drafted an AI policy for The Krehbiel Group which you can use as a starting point. You can find it here. 

https://krehbielgroup.com/the-krehbielgroup-ai-policy/