This post arose in response to an impromptu session at Worker Co-op Weekend 2025, led by Paula Mesa Macías. Although the session focused on common AI use cases and cybersecurity, it generated strong feelings in me regarding AI.

Should you have an AI policy? I believe that yes, we all should. Below, I explain why.

The absence of an AI policy already indicates a stance on AI. Usage without a policy means a free-for-all or laissez-faire attitude that could pose an ethical, reputational, data and security risk for your organisation.

As raised in the weekend’s discussion, even if your organisation is sceptical about AI, it will surely be creeping into the daily workflows of individuals (generating summaries of documents, AI-assisted search, quick ideas for social media posts, etc).

Spending time with your team to be self-reflective is an important first step in adopting a formal position on AI and necessitates examining your business’s existing usages of AI.

A frank, open, wide-ranging, forward-looking discussion seems like a good place for any organisation to start.

But how can your business have an AI policy when “AI” usage is complex and nuanced?

This is a biggie. I think we can all agree that submitting datasets of user data to an AI would be deeply problematic – if not illegal – and all of us are concerned about the mass breach of copyright by AI and don’t wish to support that.

So, how might we formulate an AI policy that encompasses such a wide range of use cases? At Outlandish, I expect we’ll ask questions such as:

  • What about coding with AI?
  • What about writing text?
  • What about image generation?
  • What about using it to search “intelligently” within documents? What about searching within your own documentation?

Perhaps one way to start the discussion about AI is to segment the use cases by medium. In other words, discuss and consider your approach to image generation. And then text generation. And so on.

Another way to consider it deeper might be to ask, what use cases are extractive? Tasking an AI to “write me a text about X” might be permissible. But “write me a text in the style of Zadie Smith” feels more extractive and directly linked to the rights of an author. So too image creation “in the style of X”.

We don’t yet have a policy towards AI here at Outlandish, and as a supplier of tech services, it is only a matter of time before we are required to submit AI policies as part of our tendering process. We plan to consider our position on AI formally very soon.

Photo by Андрей Сизов on Unsplash