00:00NVIDIA rolling out new tools designed to keep AI agents from going astray.
00:05The move comes after security breaches fueled fears about AI security.
00:09Let's get more with Bloomberg's Neil Camping.
00:11So Neil, how does this all work?
00:13Was it that easy in the end that NVIDIA could just make these tools
00:17and now they're going to stop AI going rogue?
00:22Yeah, I mean, I think this is a big deal.
00:25They've introduced this new double-layered AI security system
00:28and it says it would have prevented some of the recent high-profile breaches
00:32such as the one of Hugging Face that got a lot of attention.
00:36So by using its cutting-edge lab technology,
00:40it's really, I think, being able to put together a massive system
00:47which can control these AI models.
00:50So if this engineering solution basically ends these safety problems,
00:58I mean, this could potentially be massive
01:01given we've had so many concerns around these security breaches.
01:05So, yeah, as we get more details of the technology,
01:08I think this potentially could be a real big moment for the whole AI industry.
01:13Yeah, if it is the solve that it sounds like,
01:16then, of course it would be,
01:17but do we know yet if it works on all of the AI models,
01:20you know, Anthropics and OpenAIs and Metas and all of the LLMs?
01:26Yeah, well, that bit we don't necessarily know yet, to be honest.
01:30Now, we do know it runs on NVIDIA's own Vira CPUs,
01:35and we also know it enables users to set rules,
01:38and the software it's using is open source.
01:41So it could be potentially used and adapted accordingly into other systems.
01:46So we may not have the technical details just yet,
01:48but this could be, if it's open source, as I say, on the software side,
01:53then you incorporate that engineering into software and hardware platforms,
01:56then it could well be a big step change forward for all of the AI models.
Comments