Data Sovereignty in AI: What You Need to Know About Microsoft Foundry and Regulated Data

[birds chirping] Hi folks. Welcome to another AI briefing. My name is Tom. It's good to see you all once again. Um, today we're going to take a quick discussion about data sovereignty and in the, in the world of AI and just general sort of big data processing or data, uh, cloud-based data processing platforms, what you do with data sovereignty.

So, uh, I work a lot in the regulated industry space, and so I have to deal with an awful lot of regulated data on a regular basis, not a regulated basis. Um, and so you have to be wary about where you're gonna send this stuff. So I just wanted to raise a point I actually had with a, a chat with, uh, someone earlier today where we were discussing, uh, Microsoft Foundry. Now, for anyone who doesn't know, Microsoft Foundry allows you to deploy LLMs into an Azure environment, uh, that allows you to then process your data, so you think. Um, when I was prodding around in Foundry a while ago, uh, the thing that I noticed...

So Foundry also has a marketplace for different models, so companies can sell access to specific models into that marketplace. Now, uh, it may be that you have a infrastructure engineer, systems administrator doing the model deployment for you, and they're not necessarily aware of the data sovereignty agreements that you have with whatever project you're working on. So, for example, you may only be allowed to deploy, um, uh, certain data into a geographic region or into a certain cloud vendor in a geographic region, which is, so like quite a regular occurrence. And so be aware that when you do stuff like spin up additional models, that data isn't necessarily staying entirely within Azure. Some of those models are hosted by Azure.

Some of those models, I believe, are hosted by, uh, third-party providers, and of course, eventually your data makes your way to them, gets processed in a model, and come back. For the majority of users and use cases, it's probably of minor concern, um, unless of course you're worried about IP risk and that type of thing as well. But like obviously from a regulated industries perspective, it's definitely something to be aware of and something that you should be a little cautious of when spinning up different LLMs inside of projects like Foundry, where you're gonna send potentially sensitive PII type regulated data. Uh, that's all I wanted to say. Um, hopefully that is of use.

Uh, just something to be aware of when you are spinning these things up. Just bear in mind if you have any agreements in your contract with whoever's data you are using, make sure you hit them. I hope that's been useful. I will speak to you all tomorrow. Have a good rest of your day.

[upbeat music] Why hire when you can partner? ConceptCloud's leading engineers build your startup's prototype without the overhead. Launch faster. ConceptCloud. com.