Transcript
Differences between ChatGPT and Copilot
AI is now properly here. We’ve had ChatGPT for quite a while now and obviously recently we had Copilot launched from Microsoft. And the big difference there is that the data that Copilot is learning from is actually your own company data. It is looking at all everything you do, whether that’s files, documents, Teams, calls, emails, etc.. But importantly, one of the things we now need to consider is that level of access that it has to that information.
What are we doing around some of our sensitive information, our financial information, our HR records, intellectual property. So an example where we need to be a little bit concerned might be that we have a, let’s say, financial director that was working with us. And three or four years ago the finance director downloaded some management accounts and stored in their downloads. And then time has gone on and perhaps that finance director has left. We’ve archived all of their documents, including the downloads folder and that file with all of our management accounts. Now stored on an archive system, maybe in the cloud somewhere, buried in lots and lots of folders, maybe 14/15 folders deep. It’s very unlikely that a human being is going to go and browse through that folder structure. It’s a backup, remember, it’s an archive. And ultimately that level of information, the sensitive data inside that file is unlikely to get read by a human being.
Why AI is dangerous
But if we bring AI into this conversation, we’re effectively allowing AI to read everything we have stored in the cloud. So if I ask it a very simple prompt, like what’s the company’s overheads, then if that information is stored in the management accounts, it’s going to surface up. It’s going to be able to answer that because it’s going to take everything you possibly can and try to find some details from documents in every single file that exists and try and answer that question.
Now, that’s a problem because we may be able to layer permissions around that folder structure to stop people from accessing it. But of course, now it’s in an archive. We forgot about that file. We didn’t really remember that that file existed, and we haven’t really given any thought to the security around some of the data and sometimes we don’t even know where that data is and where that sensitive information is stored. But this allows maybe a more junior member of staff to ask a simple question to an AI system and generate potentially or expose potentially sensitive information.
So we’re not limited to obviously, the financial information, but have a think about finance documents, have a think about passports or driving licenses or HR records, people’s addresses, all that level information, intellectual property, contracts, all of that level of information could be stored all over the place in your environment. So whilst we’re encouraging you to embrace AI and we’re encouraging you to get in line with Copilot, what we’re also encouraging you to do is really look at your data, the structure, how you commission it and how you can make sure that you get the most out of it, that you don’t expose information that perhaps you don’t want to.
Join our webinar
So we’re doing a webinar shortly, really encourage you to get involved and join us on that. But there’s also some blogs and articles on our website. So please go and take a look.
