I’m in a regulated sector and there’s little guidance on AI use. Because of that, we’ve really restricted use and advise no uploading of NPI/PII until we know what the expectations are on risk assessments, policies, etc. Basically, it’s great for research, some guidance, and drafting, but shouldn’t be run on sensitive data in any form. Any decisions that are made should be your decisions and not the decision of AI. Users are trained on this. They’ve accepted the terms in our Acceptable Use Policy. We have 1 approved AI platform.

Here’s the thing - We’ve got the tech enthusiasts that reach out to the IT department regularly - “look at this new setup from Anthropic! We can automate this. This would increase productivity! We’re going to get left behind!”

“We’re going to get left behind!”
Ugh.

There’s a user that told me he’s given AI full access to his personal laptop. He has told me he personally runs bots and uses multiple platforms to write and run scripts and automations. Full access makes me cringe. Although there’s no proof that there’s “shadow AI” going on (users using unapproved tools/circumventing protection) we suspect some is going on. Knowing that these users just throw all their personal documents at these platforms worries us that we’re opening a can of worms if we un restrict AI use internally - especially since AI tools continue to evolve and we still don’t have a set expectation from a regulatory body.

It’s like any way we approach it to provide a tool that meets their expectations, it just doesn’t work. The platform we use is constantly adjusting settings as well, so it’s like every time I turn around, a new setting is on by default that shouldn’t be. It feels impossible to control.

Also - these enthusiasts are like .04% of our organization, so is the work worth it is a question that has been brought up. Even upper management feels it’s more efficient to do the work yourself. By the time your verify AI’s output or fix its mistakes, you’d be just as good doing it yourself. Management doesn’t appreciate that the work is not understood when done with AI (e.g. can you explain this line on this document? And the person can’t).

Anyone else feeling bound in a situation like this?