Did you know Splunk can provide observability for AI - including prompt, response, performance, bias detection, hallucination detection, and so much more? I recently had the opportunity to chat with Chris Doebert while at Cisco Live EMEA about the innovations happening with Splunk AI and Cisco AI Defense.

Transcript

Auto-generated captions, lightly cleaned. Speakers are not separately labeled.

Hey everybody, we’re out here at Cisco Live in Amsterdam and I ran into Chris here and we’ve been talking about Splunk and observability for AI and you gave me some really cool use cases. Can you tell us some more about that? Yeah. So, what we’re able to do within Splunk Observability Cloud is give you an endto-end view of all of your AI agents, what they’re doing, what they’re saying, how they’re figuring out how to say what they’re saying. Um, and really give you insight as granular as you want to get into that performance, into that that responsiveness, uh, really into how they’re performing for you. >> Yeah. Yeah. And you talked about the prompt and the response, but more importantly, you gave me some information about hallucinations and bias and detecting that. Can you give me some more detail there? >> Yes. So, what we’re doing is we’re using uh DPV valve, which is an open source framework to really gauge, you know, bias, uh, hallucinations, anything that that would impact an accurate appropriate answer for a prompt that someone has typed in. >> Yeah. And then so you can detect all that, observe it with Splunk, but something else interesting you showed me was this AI defense integration. Yes. What’s up with that? How’s that work? >> So we’re able to detect things like prompt injection um as well. So, not just from the standpoint of the AI agent being able to represent itself accurately and factually, but also is it being uh is is the input that it’s getting true and real or is it something like a prompt injection that’s going in and trying to subvert it in some way and get get behind it. So we’re able to really build out a framework that secures how it’s supposed to run and how it’s not supposed to run and give you all of that insight that you need into what’s happening where. >> That’s really amazing. So you got Splunk at that center of >> prompt injection, bias, hallucination, performance, all of that together and the visibility with Splunk. That’s really interesting. >> Yeah. >> Where can I go learn some more about how we’re doing all of this? >> So that’s going to be.com. Um, we’ve got a ton of information out there on uh observability for AI. We have our uh lantern article. So, lantern.splunk.com. A lot of our use cases. Uh, a lot of our white papers will show up on lantern. Um, you can get a uh free Splunk observability cloud trial for 60 days on splunk.com to try this all out yourself. >> That’s great. Thanks so much for your time. Thank you.