From Data to AI: Securing the Modern Attack Surface

Summary

GenAI has expanded the attack surface from cloud infrastructure alone to a layered ecosystem of data, models, applications, and agents, requiring an end-to-end security framework covering infrastructure, guardrails, and runtime.
Wiz integrates with Databricks to surface risks in Unity Catalog — including exposed credentials, volumes, and data access patterns — using a security graph that prioritizes toxic-combination threats that chain multiple risks together.
Over 50% of Wiz customers have achieved zero critical risk, and more than half of active users on the platform are non-traditional security personnel including data engineers and DevOps teams, enabling faster remediation.

From Data to AI: Securing the Modern Attack Surface

GenAI is reshaping the attack surface — from cloud infrastructure to a layered ecosystem of data, models and applications. As organizations build on platforms like Databricks, data becomes both the foundation of AI innovation and the primary source of risk. this video explores how to secure AI applications end-to-end across infrastructure, data, models, guardrails and runtime. Attendees will gain a clear framework for understanding AI risk across the full lifecycle and see how visibility into systems like Unity Catalog help teams understand data exposure, model usage and emerging risks across the AI lifecycle. Finally, we’ll show how Wiz integrates with Databricks to bridge security and data teams — surfacing insights, identifying risks and enabling a unified approach to securing AI from code to runtime. Attendees will leave with practical strategies to reduce AI risk without slowing down development.
Talk By: Giri Radhakrishnan, Technical Product Marketing, Wiz ; Lance Larsen, Customer Engineering Lead, Wiz ;

Chapters

FAQs

How does generative AI change the security attack surface?

GenAI shifts the attack surface from cloud infrastructure alone to a layered ecosystem of data, models, applications, and agents, creating new vectors like prompt injection in RAG applications and the need for automated red-teaming scenarios. Organizations building on the Databricks Data and AI platform must now secure the full AI lifecycle from code to runtime.

How does Wiz integrate with Databricks for security?

Wiz integrates with Databricks to analyze the security graph across Unity Catalog credentials, volumes, and data access patterns, identifying toxic combinations of risk that should be prioritized for remediation. The integration bridges security and data teams by surfacing risks in a context-rich, prioritized view.

What is a toxic combination in cloud security?

A toxic combination is a chain of connected risks — such as an exposed credential combined with broad data access — that together create a critical vulnerability even if no single finding appears severe in isolation. Wiz's risk scoring surfaces these combinations so teams can address the most dangerous issues first.

What results have Databricks customers seen with the Wiz integration?

Over 50% of Wiz customers have achieved zero critical risk on their workloads by fixing code, tightening data security rules, and reducing network exposure. More than half of active users on the platform are non-traditional security personnel — including data engineers and DevOps teams — contributing to faster remediation across organizations.

Full transcript

[00:08] One, I want to thank you all for coming here because this is the last speaking slot of the day. So, either we have people in here who don't really like the chain smokers or everyone's really excited to learn about uh how we're going to secure the modern attack surface of data and AI. I hope it's the
[00:21] second one. Uh we'll try to keep this fun and entertaining uh and get you all out of here maybe a few minutes early. I'm Lance Larson. and I lead our solutions uh engineering teams, our customer engineering teams uh at Whiz on the West Coast. I actually have a lot of
[00:33] customers that use our whiz and data bricks integration. So I'll be talking a little bit about how they're seeing value from the product today. And with me, >> go ahead, Gary. >> Gary, uh I Sorry, I thought you were going to introduce >> No, no, I would never.
[00:46] >> So I'm a technical marketing manager at Viz. Uh I'll be demoing the product. So >> Okay, that's good. So Gary's Gary's the smart one of the two. Um, we'll try to get these through these slides pretty quickly. So, just quick show of hands.
[00:59] Uh, how many people know about Whiz? Okay, a lot of people. And then same
[01:13] whiz story, uh, most recently we joined with Google. That became official in March. So, we're committed as ever to securing uh, cloud. Also, I think this really helps us secure AI. So if anything that's helping accelerate our
[01:27] our strategy and product development. Um and just two things to note that I think are really interesting across all these logos some of which are my customers. 50% of our customers have achieved what we call zeroritical. So you can think of
[01:39] a critical like an inbox of toxic combinations that's like fix this first and get it to zero. It's a lot of it's a lot of how we help people build risk programs uh in their security organizations. 50% of our customers have got that number to zero. uh whether it's
[01:52] by fixing code, battening down the network, uh tightening rules around data security, all the ways that you can fix this stuff. That's that's something we're super proud of. And the other part of this, and we'll talk about it a little bit in the presentation, is over 50% of our active users are not
[02:06] traditional security people. They're folks in data security. They're folks in development. They're folks in DevOps. Um and I think a good way to get this stuff fixed is you can't have one team bottlenecking all this. The other part
[02:18] we're going to talk about is how AI is helping all these teams get involved to fix stuff quicker. But these are two of the metrics that we're really proud of. And last, um, you know, we've we've been a leader in in cloud security. I think the premium that Google was willing to
[02:30] pay for Whiz shows um, just the the value that we were able to create. Uh, and we think with their support, we will continue to remain a leader in the space. Um, but the space is changing, right? So
[02:43] I think with AI um the whole software development life cycle not all that is different but I think there's really three key things to talk about here and you know one is how and what we build and AI is changing how a lot of these
[02:56] applications are built right a lot of you're hearing a lot now just how many applications are built uh with AI code assistance as one example and the data flows are also changing with AI you know what what are these applications trained
[03:09] on what knowledge bases do they have for rag or for context retrieval. And then I think also and this is the more interesting one probably of the three is just how risk emerges. We're seeing with these frontier models, you know,
[03:22] typically it'd be like a high severity alert uh that's publicly exposed that gets compromised, but models like, you know, Mythos uh and some of the other frontier models are actually able to chain like low and medium vulnerabilities together in ways that
[03:35] we've never seen before to do these attacks. So I think um how and what the the way the data is used and how risk emerges, these are the the ways that we're thinking AI is really changing uh how you need to think about posture and protecting your cloud environment. Um
[03:48] and here's a good example. I love showing the security graph because it's visual. Um I once had to do a talk in Japanese. I don't speak Japanese. So I literally just presented the product uh in Japanese localized language but with
[04:00] the security graph and everyone in Japan got it. So uh picture says a thousand words. And what I'm trying to say here is the the layers of code and infra and runtime that work together. Um I think
[04:12] code is kind of becoming the most interesting one especially with AI. We use a lot of AI in our product and now we're able to understand code intent and understand what an application is doing and that's really interesting context uh
[04:24] in security. The infrastructure is still a super important point in that story. um you know I remember when deepseek was online it was actually compromised by a whiz researcher who just had a database key or something that was uh they were
[04:37] able to retrieve in plain text and dump the whole thing right so in infra is also important and I think the other thing is runtime is becoming increasingly more important the way these AI applications are behaving um people don't really know exactly you
[04:51] know what they're going to do so being able to have runtime signals of when an agent might be behaving maliciously or doing something outside its intended behavior these things all come together to secure the AI stack and we're still driving this through the whiz security graph that our customers know and love.
[05:06] Um, and I think again, you know, from the data and application standpoint, um, not a lot of this changes. It's still the same security graph, but I think specifically now we're just seeing a different set of applications deployed
[05:18] on top of Whiz. And to my point about understanding the code intent, um, whiz is getting to Whiz is getting more intelligent around like, hey, if I want to red team a specific application, a
[05:30] rag application might behave differently than an AI application, can I learn about the behavior of that from the code that's written and actually automatically pentest that in an intelligent way? That's something that's new, really exciting, like combining AI
[05:42] SAS with automated red teaming to kind of do this around more of the application layer in intelligent way. I think this is the stuff that I'm the most excited about and it's the stuff that customers say, "Wow, you know, I haven't really seen that in a closed loop." That's really neat.
[05:55] Um, sorry, I hate slide transitions. I asked for them to take these out. I always miss these. Um, so the so you know, putting that into a risk framework, right? Because I think you're sort of working with different teams in how to think about this problem. We're
[06:07] still starting a lot at that infrastructure level and infrastructure enables access and that's where a lot of the permissions and you know underlying plumbing that makes the AI work that stuff still has to be secure right and
[06:20] that could be the weakest link I think that's well understood the other parts like models you know driving decisions whiz is adding a lot of intelligence just for people to understand hey which models are running in my environment what might be running from hugging face
[06:32] are there older models that you know might have pietorch and code execution problems. We ran a project with a customer and that you know CISO found out that 30% of the models in production were on security's list of models that
[06:45] were uh unwanted or posed a security risk. So from an inventory and visibility perspective models are still driving the decisions. We want to make sure good models are being used and guardrails are in place. The data is really shaping the AI behavior. So
[06:58] systems now are relying on data in a in a much different way for context, for retrieval, and for decision-m. And a lot of people try to trick agents into getting data that they shouldn't have. And I think that's another um really
[07:10] interesting part of the story. And we have some great customer examples of where we've helped people detect malicious agents with too much agency doing too much stuff outside of data boundary. And the last is really the application and and you know layer. And
[07:22] whiz has added this um agentless scanning engine that now we can see where agents are running in the environment and actually understand what tools they have access to. Are they right agents? Can they execute code? And
[07:35] then sort of build out more of a a model of what the agents are doing in the environment. And a lot of people that we talk to say I just don't have this visibility. You know, we have a governance process, but once stuff's in the environment, um you know, these are
[07:47] some of my gaps. So whiz is adding value for all this and it's a it's a important part of the overall story. Um but I think like when you look at it together, if you have just different tools solving each piece, it can be hard to stitch the
[08:00] story together. But if you have a tool or platform where you can do all of these things in one, then you can really start to make a difference in working on that AI posture and bringing that AI risk down. Um, and I think this is
[08:12] probably no surprise to people, but customers, I think the token spend and the token maxing and all that's kind of legit. I think we're still trying to find our way and like realizing the value of what we're doing with AI, but
[08:25] the adoption speaks for itself. You know, we're seeing more customers than ever. Um, self-hosting models, I think that's a really interesting insight. Uh, using AI coding co-pilots, that probably shouldn't surprise anyone. MCP, everyone
[08:37] hates it or loves it, but it still seems to be out there in good form. And uh managed AI agents. This is an interesting one from the cloud providers. Um things like Gemini enterprise platform uh as one example. Like we're just seeing more investment
[08:50] in some of these platforms that are starting to mature. So I think for us, we still want to help you protect everything you code. You know, make sure from an SDLC perspective, it's tested, build, and certified. And then when it's deployed and running, can we provide
[09:02] that? So you can kind of think about whiz doing those four pillars across all of the software development life cycle for all your teams and that's really what we're trying to do um with whiz some of those questions right so you know security might uh we're going to
[09:15] show the data bricks product so you're going to see some of these examples but which data bricks exposures are actually uh exploitable in the cloud so you might have a bunch of data bricks keys or other you know workloads consuming the stuff but whiz might be able to say hey
[09:28] here here's data bricks credentials that are at most risk and they're in these uh sensitive applications that you should fix. For people that are actually building all the code, um they might want to understand, you know, which permissions, configurations, and
[09:40] dependencies could reduce risk. So they could work against their key metrics, data and security class of questions. You know, where is my sensitive data exposed? Who has access to sensitive data? Um and the security graph really can answer all these questions and
[09:52] operationalize it. My mom asked me once, you know, what is whiz and why does Google want to pay so much money for it? And I just told her Whiz answers questions for people that nothing else
[10:04] does, right? So when these teams come together and they ask the important questions to them to do their job, they can go to Whiz and find the answer with high fidelity and be confident in that. And I think that's why we've done well
[10:16] helping people drive the risk down in their programs. Okay, so that's kind of the general whiz pitch. And I think data bricks is a critical surface. We have a lot of shared customers who use data bricks and
[10:28] I think it's a really important part of this connected AI ecosystem. So uh for us you know data bricks is a place where identities data AI applications and cloud infrastructure it all sort of like
[10:40] intersects um and understanding risk requires context across the entire environment right and that's not just cloud. I think we're seeing data bricks has done an incredible job of really
[10:52] coming with a good AI strategy bringing that to the customer. So we want to bring the same context that's really useful in whiz to data bricks. So you kind of see that um come together but what that actually might look like. So
[11:04] over here um kind of same you know whiz security graph and I think this is some application that I don't know what it's doing for the sake. it doesn't really matter. But what we're saying here is there's like a critical, you know, vulnerability on this machine and you
[11:17] know this machine has an actual data bricks access token that we validated that has access to this service principle that could go and have access to sensitive data. Right? So um it's the like data bricks is is totally
[11:30] integrated to just work in the security graph. So we can take in that context, run it through all of our engines, and then from a discovery perspective, um we can get that information, we can put the
[11:42] whiz context over it, and then we can go to the security or data team and say like, hey, here's all the data bricks credentials that are outstanding. Of the 1,000, these 50 are on sensitive applications with risk. Just fix those.
[11:55] And you can focus on 50 things in a twoe sprint if you're trying to close down risk. you're unlikely to ever find a enough a team enough time to deal with a hundred thousand or 10,000 security findings that no one has time to do,
[12:08] right? So, it's just a great visual story. Looks good in English, Japanese, anywhere else. All the teams can just look at it and be like, "Okay, we don't have to argue about this or send a thousand Slack messages to one another. Let's just go fix the thing." Um, and
[12:22] this slide kind of just this is my main point, though. It's taking hundreds of thousands of findings and trying to figure out, you know, what what's the hundred things or what's the 50 things or 10 things that we have time to do right now, right? We've always been really good doing that at Whiz, in my
[12:34] opinion. Now, we're bringing data bricks into that same story so security teams and data teams can work closer together on this story. Um, and I think in the AI,
[12:46] has anyone heard about the Whiz red, green, blue agents? I saw a lot of hands up on Whiz, but I'm just really curious if people have heard about our AI stuff. just a couple folks. Okay, so not a lot. So this is actually really interesting. Um,
[12:59] so we have a red agent now that is uh it's basically like an automated pen tester. So it goes and it finds like it'll it'll take account of all these exposed applications on the internet and
[13:13] it'll start to use all the context it has about like what kind of coding application it is. We even have a code AI SAS engine that uses AI to come up with vulnerabilities in the code that it then sends to that agent to test. It's
[13:27] actually really cool. It's one of the coolest parts of our product. So, you can think of just how much work pen testing is. But Whiz is using that entire context graph to do the most intelligent purpose-built pen testing. Um, and we've had customers, this has
[13:39] been in preview now for a month or two, and we've had customers find hundreds of validated issues like large financial companies with uh transactional databases that were able to be dumped on
[13:51] the internet. We had another customer with a a chat, you know, data bot where our red agent found that uh it could actually be tricked into returning any information for any customer uh unauthenticated, including like social
[14:04] security numbers and other PII. So real things, right? And I think um we've always had the ability to pentest, but with AI, people are able to attack us faster than ever. So we need to be able to beat them to the punch. And if we
[14:17] have all the data, we should be able to beat them, whether it's on data bricks or the cloud or any other platform. That's the defender advantage, that data. Um so that's the red agent. It's really cool. Uh we have a blue agent that works um basically for real-time
[14:30] threats. It can kind of do like a quick, you know, sock analysis, blast radius, all that stuff. We had one customer um their investigations went from like two hours to 30 minutes with the blue agent. All that manual work that the teams
[14:42] would do in the war rooms when they would get together, you know, a bunch of that was automated. They could show up and do the last mile. That saved them, you know, 75% of the time when doing these investigations. So that's like a super smart sock analyst that's work on
[14:54] the data. The green agent is also really cool because it takes all the stuff in the platform and it actually uh suggests automated code fixes uh that you can then ship to coding agent or manually fix. So it'll take all that information
[15:08] and it can open like a pull request intelligently with like here's the problem, here's the issue. It can call a coding agent into that workflow. It can give that information to a human. So that's really the closed loop around fixing stuff. So, um, the the way all
[15:23] these agents come together is it's it's basically we're calling it like machine speed defense, but it's can can you fix as fast as things are being attacked? And I think the remediation problem is the hardest one. It helps with
[15:35] ownership. It helps suggest intelligent fixes. Um, and we're just seeing people uh like we have a lot of customers where we build vulnerability management programs with them and we're just seeing a much much higher accelerated rate of
[15:47] fixing when we have the green agent in there determining who owns the fix, what they need to do, um, and showing like the validation story and proving that it was something that we should actually fix. So, I don't know if we're going to cover too many of the agent stuff in
[16:00] this demo, but for everyone who knew Whiz, there were not a lot of people that kind of were familiar with this. So, definitely go check out the red, green, blue agents. I think this is the most interesting part um of the Whiz platform.
[16:12] So I think uh finally and this is some of the stuff that we're going to show next um basically again back to my original point whiz helps people answer questions understand what exists map it on to sensitive data and secrets
[16:24] understand who can access to what and now we're just bringing all that to AI and so when you look at these examples in the product this is what you're going to see I'm going to skip a couple of these slides because we're going to show them in the demo um but again just f of all the stuff here that's new for the
[16:37] whiz folks red blue green agents um at least the our stuff, not the data bricks stuff um that we're going to talk about in a second. This would be the one thing I would go check out coming out of this. It's brand new. Customers love it. Um
[16:50] it's really interesting and I think putting all of our platform together, all the products and putting this agentic security model around them, that's the the most interesting that thing that we're doing and that we're working on through the remainder of this year. Okay. Um
[17:04] what does this say? Okay, we talked about this enough. It's demo time. We did enough slides. All right, take it away, Gary. >> Thanks, Lance. Uh, all right. So, I'll definitely show the agents
[17:17] uh during the demo. Let me just change to my browser. All right.
[17:34] Okay. So, we're here at Data Bricks Summit. I'm assuming teams are building with data bricks. Show of hands, how many of you have vioded? Nice. So, we all know that you know people are wipe coding using maybe lake houses with
[17:48] uh with data bricks and CEOs are asking hey build me a chatbot. I want to see what's going on within the company. Uh code is pushed uh into production just couple of days and then the CEO is out
[18:02] you know traveling somewhere. He wants to he or she wants to access the chatbot anywhere they go. Uh and then you start doing it. It exactly happened at Viz. You know our CEO wanted to do this. He wanted to look at customer information.
[18:15] So what you see here is actually that data bot. This was vipcoded in a week. Uh and you know things like customer data revenue and all these things you you go you know ask questions on this
[18:27] chatbot and the CEO gets uh all the responses they want. But no one was asking is this secure? How did we build this? That's where viz comes in. Um so
[18:39] Lance was talking about all the AI stuff. So let's see here what it does. So we start with inventory, right? That's how you get visibility. All you see here just like the cloud infrastructure that viz is known for uh
[18:51] in terms of visibility. The same happens at the AI layer. So on the menu on the left here, you see all the agents, the models, guardrails, MCP servers, everything here. is from your AI layer. So I spoke about
[19:04] a chatbot, we called it the data bot and you see here we show it as a as a service catalog. So that's how you map you know whatever you build using your AI uh in terms of services. So let me
[19:17] just click on this uh data board and see what it shows. So the team that built this data board army uh I don't know if you can see the other two people. Yeah. So our founders Ami, Roy and Yunon built
[19:32] this data bot and below that you see the service architecture which shows how this data bot is built on the uh you know in the cloud. So it uses AI tools. There is an agent uh there are a couple
[19:44] of API endpoints and uh these AI tools show what this I don't know if it's too small for you to look at but yeah it can execute code it can query databases uh it can list tables. All right. So that's
[19:58] the data bot. Uh but then what is uh whiz doing with all this information? So we show you all the risk issues, right? We talk about risks. Uh and in this particular issue, it's basically a
[20:11] combination of different findings and it's all about uh combining these different findings. So you see that particular data bot uh there is an agent. It's externally exposed with authentication bypass. And how did we
[20:24] prove that? So Lance was talking about red agent. So here's a red agent. What it actually did is it went and checked this data bot to see if it can you know bypass authentication. So it just gave
[20:38] it random uh you know bear token without any type of uh checking if it's valid tokens or not and it just you know went in and it could access uh you know PII customer information. It is extracted uh
[20:51] financial data. So this is not good right? So there is real validation. It's not just a theoretical risk saying that hey your uh data board is not secure. Uh and you can see so how this data board
[21:04] is built is you can visually see it through this architecture. So there's an EC2 instance and it's hosted the data bot here is hosted on that particular EC2 instance and you also see the the AI
[21:16] tools that's part of the uh data bot. And then moving to the left, you can actually see that if an attacker has access to this data bot, they can actually go and go into the S3 bucket. They can go and reach sensitive data
[21:30] like financial credit card information, PII and uh you know other sensitive data. So that's all the issues that you know whisk can unlock
[21:42] data bricks here. Let's assume that this particular data bot behind the scenes you're using data bricks viz. So I spoke about the visibility and uh you know in all the inventory that
[21:54] viz can uh detect in the cloud infrastructure. Let's let's just look at data bricks. So we actually have a unified connector. It's probably the only security tool right now that has a connector for datab bricks. And what it
[22:07] can do for you is inventory or show you the inventory of everything that's part of your data bricks. uh deployment. So everything starting from your workspace, your Unity uh catalog, I guess that's
[22:21] that's what it's called. Yeah, the Unity uh uh catalog uh your tables, your volumes, all the data. So Viz is scanning all of this and it's just you know minimal effort for the user to get this uh done. You just start with the
[22:35] connector and then Viz is showing you all of this. U so this is the inventory of the cloud resources. you can just filter uh for data bricks and it shows you this and then I'm going to talk about you know three types of risks here. So we saw the risks with uh the AI
[22:50] security module but then specifically just for data bricks we're going to look at three things. One is lateral movement to admin uh the second is u credential compromise and the third is sensitive
[23:03] data exposion. So what you see here is all account users in data bricks with later actual movement to admin. So this is an issue that whiz has uncovered. So you see here if your data bricks administrator has set up your uh
[23:16] you know workspace in a way that all users are grouped into one uh you know one group they can actually there is a if you see at the if you see the service graph you see that there is a lateral movement finding. So any user part of
[23:29] this group can escalate themselves to an admin role and this is not you know best practice for uh identity right u you can see that they are part of this uh datab bricks account and there is also a
[23:42] remediation uh this is the green agent that Lance was talking about so I've covered the red and the green uh hopefully in the next uh issue I can show you what the blue does but then the green agent is telling you how to you
[23:55] know remediate or fix this particular problem. Um, next you go to data findings. So, viz is also scanning all the volumes, all the tables part of your data bricks
[24:07] uh setup, right? I'll just take a simple example. This is a readonly uh volume and whiz is seeing all the information out there. So you can see there's about about 4,000 odd email addresses uh date
[24:21] of birth uh you know of users, employee ID, social security number, all this sensitive information. So it's hosted on you know it's present on this particular volume. It is sensitive data
[24:35] has to live somewhere but then what is the risk here? We'll see uh what happens next with the next issue where this particular uh volume lives. So there is an other issue and you know I don't want
[24:47] people to think that hey you said single pane of glass you're clicking through four different tabs I'm just doing this because the internet is really shoddy here I didn't want to take risk by and delay this even further. So in this issue what you see is all account users
[25:01] you saw the group of you know user account in one particular group and this one this issue says all account users in data bricks with data access to sensitive data. I just showed you the read only volume that had you know those
[25:15] 4,000 email addresses and uh date of birth information. You see that right here you see the uh account users group. It directly has access to this readonly volume.
[25:28] And if you click on that particular volume it's going to show the same data again with all these email addresses and so on. Uh so this is again one more uh finding that viz has done with uh you
[25:41] know with the datab bricks connector. You'll again see a recommendation here to restrict data bricks uh accountwide data access to sensitive data. Uh the final one I'll show you this is like the
[25:54] nail in the coffin where uh publicly uh readable bucket contains clear text SAS keys. I think Lance also briefly mentioned about you know this could be a use case where you're just putting out clear text keys inside uh buckets. Uh
[26:09] let's just assume you know since this is uh internet facing and publicly available it's not even a an issue of there's there's not even social engineering involved here. Anyone who's just scanning for this public bucket uh
[26:21] they get direct access to this uh storage account with uh sensitive data. So those are the three use cases I want to touch upon with data bricks. Uh so what we see here is it's all three
[26:34] layers right viz is not only about cloud infrastructure layer now it also has coverage for AI layer and also the data bricks connector combined. So that's my demo. >> Gary can just I just want to call out
[26:46] one thing. Can you go back to the first issue? >> Okay. >> The data bot one. >> Uh >> yeah. Okay.
[27:05] >> And then can you go to the remediation tab really quick? >> Oh, yeah. I think >> this is a good one. So, like show this uh step four, just pick any of the coding fixes. >> Okay. >> Yeah. So
[27:19] basically when I talked about actual understanding of the code so when we're doing all this we're actually now you can take the the the specific things that would solve
[27:31] this problem and send it directly to a coding engine. We also have like a times like workflow thing now in the product where you connect this agents to your agent. So if you have a superfixer agent you can have it communicate with our
[27:43] agent to get this context and go fix stuff. But I think this is really powerful because to the point that was just made, we're not that far away from like you built a data chatbot. It could be getting data from data bricks. It
[27:55] could be exposed on the internet. Whiz could come up with its own AI SAS finding, like not a static SAS rule, but literally looking at the code and detecting something, sending that to our red agent that knows exactly which tools
[28:09] potentially expose that code, exploiting it, and then coming back to a fix. All without a single human being doing anything. And that could be on the cloud
[28:22] or it could be on the cloud connected to your most sensitive data in data bricks. And I think so we showed you a lot of different issues because we wanted to show you the coverage and the supportability of like data bricks because there's a lot going on there and
[28:34] we're very committed to having parity on the platform. But like if you just kind of think about that whole life cycle um all the way from you know when that
[28:46] application's built packaged put into production. This is when we call machine speed defense or the defender advantage or using all the context you have. The attacker doesn't have any of this. All
[28:59] they have is an orange dot that has a port exposed on the internet. then they have to do all the work to figure out what kind of application is, how to exploit it, you can do the exact
[29:11] opposite. You have all the data and if you can feed that into an AI um security powered platform, then you can drive amazing outcomes. And now we're just bringing data bricks natively
[29:24] into that story with as much par for that platform as humanly possible. So, thank you everyone for coming.

Learn more about the Databricks Data and AI platform.

The information provided herein is for general informational purposes only and may not reflect the most current product capabilities or configurations.