Loading summary
Transcript1 lines
- [00:00]
A
Welcome to the IA on AI Podcast, part of the Audit Podcast Network, where we bring you weekly updates on AI from the internal auditor's perspective. Here we go. From computerworld.com Keep AI browsers out of your enterprise, warns Gartner. Analysts at Gartner warn that emerging AI enabled web browsers like OpenAI's Atlas and Perplexity's comment pose unmitigated security risk and potential irreversible untraceable data leaks. They recommend that enterprises ban these browsers entirely. I think if you are on your AI governance committee, it should be raised to those folks as well. So if you think about like a browser that has AI or even agentic services as part of it, it's going to be able to see the web pages you're at, your emails, internal content also, and it's going to process that and potentially expose that sensitive information. We've talked about prompt injections on the show also, but those that are. That's maybe still a new term. That's basically where you have the AI browser and it is autonomously doing its work through the agentic services that it has and maybe it accidentally clicks on a phishing link or something to that effect. There could also be hidden text somewhere on the page that's telling your browser what to do, as if you were writing the prompt. So the way, the same way you might prompt it to do something that could be hidden somewhere on a website or something like that and the browser is going to pick up on that and take it as a prompt coming from you. Same way if you know about SQL injections, something similar to that effect. I know we made the recommendation when these first started coming out. Just don't download them for sure. On work laptops I don't have it on a personal. I'm not going to have it on the work one either for the foreseeable future and actually to that point on the foreseeable future. This article goes on to say emerging AI usage control solutions will likely take a matter of years rather than months to mature. I'm usually not a huge fan of just like hey, let's block everything so that nobody can have access to it. But I think I would be willing to make the exception here if I was the CTO or someone in that role. So I would check see if you are able to download these and if so I would highly recommend you can use this article as further proof. I'm sure you went to your AI governance committee or whoever and said Trent Russell said don't do it. They'd be like, I don't know who that is and don't care. But maybe if it's coming from Gartner they will be more prone to listening. From Reuters OpenAI warns new models pose high cybersecurity risk I don't think this is like a huge shock to anyone, but they did send the warning that with their next gen AI models they could pose a high cybersecurity risk as they might autonomously develop zero day exploits or aid complex hacking operations. AI driven cyber attacks are going to be insane. I think at this point the best thing we can do from the internal audit perspective is to evaluate with incidents response plans. Make sure we're good to go from there. Thank you for listening and be sure to follow the link to greenskiesanalytics.com in the show notes and schedule time to see how green skies can make the hype of AI a reality in your internal audit department.