WEBVTT

NOTE Sentence-level transcript of https://www.youtube.com/watch?v=1EZdpEhwmNc

NOTE One cue per sentence. Cue ids are the line anchors on /transcripts/1EZdpEhwmNc.html. A cue ends where the next begins, or 2 s after its last word.

s1
00:00:01.309 --> 00:00:03.309
[music]

s2
00:00:12.840 --> 00:00:14.000
What's up, everyone?

s3
00:00:14.000 --> 00:00:18.760
Good to see you all here in the very first ever security track at the World's Fair.

s4
00:00:18.760 --> 00:00:31.560
Um pretty exciting day, honestly, because uh like all of you, I genuinely love this stuff, and having looked through the agenda for the speakers we have today, um it's going to be absolutely mind-blowingly

s5
00:00:31.560 --> 00:00:33.040
useful and fun information.

s6
00:00:33.040 --> 00:00:45.120
So, hopefully if you stick around all day, by the end of the day, you'll be able to leave here, go back to your hotel rooms, and build some genuinely cool software, hopefully without humans in the loop, because that's the ultimate goal, right?

s7
00:00:45.120 --> 00:00:49.960
Like, how do we build truly safe autonomous software at scale?

s8
00:00:49.960 --> 00:00:52.040
And it's not something easy to do.

s9
00:00:52.040 --> 00:00:57.960
So, with that being said, I'm very excited to welcome my my good friend and colleague, Manoj Nair.

s10
00:00:57.960 --> 00:01:01.960
He is Sneak's Chief Innovation Officer and CTO.

s11
00:01:01.960 --> 00:01:06.160
Um before Sneak, he was the Chief Cloud Officer at Convo IT.

s12
00:01:06.160 --> 00:01:08.240
He founded and ran HyperGrid.

s13
00:01:08.240 --> 00:01:12.960
He did product and security leadership at HPE, Dell, and RSA.

s14
00:01:12.960 --> 00:01:15.600
And he's got something like a dozen patents to his name.

s15
00:01:15.600 --> 00:01:18.280
So, he's kind of a legend in the space.

s16
00:01:18.280 --> 00:01:21.880
Uh he also personally had a hand in curating this entire track.

s17
00:01:21.880 --> 00:01:27.840
So, if you like the talks today, please go up and say thank you to him after, but if not, then just don't blame me, basically.

s18
00:01:27.840 --> 00:01:30.520
Um but yeah, welcome to the stage, Manoj.

s19
00:01:30.520 --> 00:01:31.700
Woo!

s20
00:01:31.700 --> 00:01:33.400
[applause]

s21
00:01:33.400 --> 00:01:36.160
Thank Thank you, Randall.

s22
00:01:36.160 --> 00:01:37.960
I was not expecting a bio.

s23
00:01:37.960 --> 00:01:38.840
Hi, everyone.

s24
00:01:38.840 --> 00:01:45.600
Um really appreciate uh you all joining here um right after those great keynotes up front.

s25
00:01:45.600 --> 00:01:57.280
Uh I'm Manoj Nair, and uh I have the uh pleasure of leading an amazing team that is helping secure about, you know, 5,000 enterprise customers around the globe.

s26
00:01:57.280 --> 00:02:03.800
Uh so, I get to look good about all of that, but some of what I'm going to show is real data from those customers.

s27
00:02:03.800 --> 00:02:11.920
Half of Fortune 2 Fortune 500 runs on Sneak, and some of the data is, you know, from those learnings.

s28
00:02:11.920 --> 00:02:17.440
But before that, I I also want to like talk about this, you know, the title of the talk was cutting through the AI fog, I think.

s29
00:02:17.440 --> 00:02:21.600
But the way we do that is by having tracks like this.

s30
00:02:21.600 --> 00:02:30.040
So last year we stood here, there were 3,000 people at the AI Engineer was there, and you know, it really felt like security was missing in the room.

s31
00:02:30.040 --> 00:02:37.240
And thanks to Swig and amazing partnership with the AI Engineer organization, we created the AI Security Summit in partnership with them.

s32
00:02:37.240 --> 00:02:46.320
And we're creating this track, so it is really good to see this and all the great speakers who are going to, you know, talk um here today with some fantastic knowledge.

s33
00:02:46.320 --> 00:02:50.320
But that is in our mind, like that's how we cut to the fog, right?

s34
00:02:50.320 --> 00:02:53.120
Security needs to be very much part of the room.

s35
00:02:53.120 --> 00:02:59.560
We're very passionate about it not slowing things down, but really that's how you build trusted systems.

s36
00:02:59.560 --> 00:03:08.560
So, one thing you're going to hear from from me quite a bit of, you know, in this like, you know, what if you take one thing, it's this notion of

s37
00:03:08.560 --> 00:03:20.519
our learning from this real-life data and working with the biggest frontier labs in the world and the biggest companies in the world is this concept that has really been, you know, not questioned in security

s38
00:03:20.519 --> 00:03:23.200
before, but it is being asked now, right?

s39
00:03:23.200 --> 00:03:27.200
Can, you know, the generator and the validator be the same?

s40
00:03:27.200 --> 00:03:31.680
And our point is, you know, in some of the data you'll show for all kinds of reasons why not, right?

s41
00:03:31.680 --> 00:03:37.040
And and almost like if you know Sneak, don't think about the supply chain security company that shifted left.

s42
00:03:37.040 --> 00:03:46.640
Like a lot of what you'll see is the last 18 months of what we have been doing with some of these very large enterprises in in adopting, you know, these complex

s43
00:03:46.640 --> 00:03:52.080
uh systems that are we're all in joy and we love, and what What are we learning from that?

s44
00:03:52.080 --> 00:03:57.960
Um, there are three problems that, you know, we hear when we talk to these customers.

s45
00:03:57.960 --> 00:04:01.640
And I want to share those and I want to like kind of show some of the data behind that.

s46
00:04:01.640 --> 00:04:12.640
Um, but, you know, fundamentally, it's it's really looking at this as um you think about this room and everyone that you support, you know, you're building fast at the frontier.

s47
00:04:12.640 --> 00:04:19.359
The question nobody is answering is, can you trust what your agents just shipped and how they did it, right?

s48
00:04:19.359 --> 00:04:30.919
And so if you'd like really take, you know, the gist of and I I have the joy of talking to all of these customers a lot, this is pretty much every like one, two, or three

s49
00:04:30.919 --> 00:04:32.080
or all of them.

s50
00:04:32.080 --> 00:04:34.680
It's really the con- conversation that happens.

s51
00:04:34.680 --> 00:04:41.360
So, autonomous attacks, like, you know, it's not mythos, it's not, you know, I know I said the M word, but sorry.

s52
00:04:41.360 --> 00:04:45.160
Uh, you know, it's it's not, you know, uh, GPT 55 Cybers.

s53
00:04:45.160 --> 00:04:47.880
Like, we're seeing that you can have these attacks.

s54
00:04:47.880 --> 00:04:51.080
All the frontier labs have been used in automated attacks.

s55
00:04:51.080 --> 00:04:54.400
You can do that even without having frontier models.

s56
00:04:54.400 --> 00:04:57.800
We have shown it, so have the attackers, unfortunately.

s57
00:04:57.800 --> 00:05:02.040
With good context and good harness, now you have an attacker that never sleeps.

s58
00:05:02.040 --> 00:05:02.840
What does it do?

s59
00:05:02.840 --> 00:05:10.120
The fundamentals of things like application security that was already something we tried to disrupt for 10 years have completely, you know, gone away.

s60
00:05:10.120 --> 00:05:17.800
Like, oh, you cannot have contextual risk management and say, I fixed my criticals and I fixed my highs and I'm pretty good because everything else is too hard.

s61
00:05:17.800 --> 00:05:22.400
No, you can string low vulnerabilities and and, you know, create exploits.

s62
00:05:22.400 --> 00:05:26.040
You can do it without a lot of a harness with the mythos class model, as we've seen.

s63
00:05:26.040 --> 00:05:29.800
And and but with a little bit of, you know, effort, you can do it with everything else.

s64
00:05:29.800 --> 00:05:39.600
So, so that's a big big, you know, sea change in how, you know, we are seeing our customers actually react real time.

s65
00:05:39.600 --> 00:05:46.000
And then you go, wow, a thousand enterprises spend over a million dollars for, you know, a cloud code roll out.

s66
00:05:46.000 --> 00:05:47.600
So, this is the answer, right?

s67
00:05:47.600 --> 00:05:50.240
Well, mhm maybe.

s68
00:05:50.240 --> 00:05:52.840
Let's let's let's pull that string a little bit.

s69
00:05:52.840 --> 00:05:57.920
The There are existing classes of problems that are getting worse.

s70
00:05:57.920 --> 00:06:01.880
The quality of code is unfortunately worse than human-generated code.

s71
00:06:01.880 --> 00:06:06.120
It's not like humans we I'm an engineer like I don't think I wrote perfect code, none of us do.

s72
00:06:06.120 --> 00:06:08.480
So, but it is actually a little worse.

s73
00:06:08.480 --> 00:06:11.560
And so, well, if that was the only problem, that's okay.

s74
00:06:11.560 --> 00:06:12.840
But what about the environment?

s75
00:06:12.840 --> 00:06:21.320
All the things that we find as magical are based on skills and MCP servers and all these things that we want to share and use.

s76
00:06:21.320 --> 00:06:30.800
And those are intentionally or unintentionally both, you know, being poisoned and, you know, malware's injected in there and we'll show some of the research on that.

s77
00:06:30.800 --> 00:06:32.280
And then the behavior of the agent, right?

s78
00:06:32.280 --> 00:06:39.040
And so, so we're having these these, you know, new patterns of problems compounding on top.

s79
00:06:39.040 --> 00:06:44.000
And then who here doesn't want to become a an AI company or in any room in the world, right?

s80
00:06:44.000 --> 00:06:51.200
Every company every board's like we're going to transform our business, our workflows, or you're starting brand new, you are fully agentic.

s81
00:06:51.200 --> 00:06:57.320
And when you build with agents and models, that's an entirely new threat surface that was not part of the prior one.

s82
00:06:57.320 --> 00:06:59.760
And so, these are really the fundamental problems, right?

s83
00:06:59.760 --> 00:07:03.080
And And this is, you know, I want to share some of the real data.

s84
00:07:03.080 --> 00:07:07.680
This is 4,800 plus customers in the last year.

s85
00:07:07.680 --> 00:07:14.440
Their actual backlog quarter over quarter is like 108% more backlog.

s86
00:07:14.440 --> 00:07:17.480
So, remember that what I said about attackers?

s87
00:07:17.480 --> 00:07:20.200
It's not just the existing vulnerabilities.

s88
00:07:20.200 --> 00:07:24.160
And unfortunately, there's millions of them if you're an enterprise of any size.

s89
00:07:24.160 --> 00:07:32.120
It's the fact that we are growing them despite the best agents, despite things that we have done and the industry has done.

s90
00:07:32.120 --> 00:07:35.880
This trend is real and it's it's not good.

s91
00:07:35.880 --> 00:07:40.000
And, you know, you have novel exploits, but they're not novel.

s92
00:07:40.000 --> 00:07:48.160
Like the light LLM exploit, you think about like it's really taking existing vulnerabilities in the new surface and and chaining them together.

s93
00:07:48.160 --> 00:07:53.440
And so, they create a much bigger blast radius right now at the pace at which work's being done.

s94
00:07:53.440 --> 00:07:57.320
And yes, speed is a big part of this, but it's not just speed.

s95
00:07:57.320 --> 00:08:10.880
And you know, just this um last week we had the Five Eyes uh you know, uh these are the the the Western world's uh intelligence leaders talking about AI will bypass cyber security systems

s96
00:08:10.880 --> 00:08:11.880
in months, not years.

s97
00:08:11.880 --> 00:08:15.800
Now, I don't share you know, share this I hate to have the scare tactic.

s98
00:08:15.800 --> 00:08:17.080
It's just a fact.

s99
00:08:17.080 --> 00:08:23.120
It's getting ready for you know, whatever is coming and we have already seen what's coming.

s100
00:08:23.120 --> 00:08:25.200
It's just not widespread enough.

s101
00:08:25.200 --> 00:08:33.760
And so, trying to like you know, chase systems and like the specific model will be the cure for all this is not the way is our point.

s102
00:08:33.760 --> 00:08:40.240
The data on the second pain point, the untrusted output environment and behavior of agents.

s103
00:08:40.240 --> 00:08:41.719
And these are again facts.

s104
00:08:41.719 --> 00:08:44.080
These are benchmark data facts.

s105
00:08:44.080 --> 00:08:50.880
All of them the QR codes at the top are are if you want to like go check out the studies and the research behind it.

s106
00:08:50.880 --> 00:08:55.680
This is you know, the amount of vulnerable code coming from the latest models.

s107
00:08:55.680 --> 00:08:58.200
Plus the skills, there's toxic skills.

s108
00:08:58.200 --> 00:09:08.640
We found the research the seminal research around how skills a third of them or more than third of them in all of the skills, not just you know, open claw,

s109
00:09:08.640 --> 00:09:17.640
clawed, code acts, these skills actually have malware and they have vulnerabilities that are being Three lines of English are able to now bring a system down.

s110
00:09:17.640 --> 00:09:20.320
So, you have to really understand the intent behind it.

s111
00:09:20.320 --> 00:09:23.640
And the MCP servers, how do you connect to enterprise data?

s112
00:09:23.640 --> 00:09:27.120
This is great protocol, very little security built in.

s113
00:09:27.120 --> 00:09:35.480
It's getting better, but the foundations behind it is you know, this is the GitHub MCP server exploit that we highlighted to the year a year ago.

s114
00:09:35.480 --> 00:09:37.760
And what did some of our customers do?

s115
00:09:37.760 --> 00:09:47.480
Immediately shut down all MCP servers then they figured out that all of their dev screens and then how do you actually go back safely enabling things that are very powerful.

s116
00:09:47.480 --> 00:09:51.960
And then on the behavior, everyone knows the pocket OS example, right?

s117
00:09:51.960 --> 00:09:57.760
What I have is real data in our own environment in our Fortune 100 customers.

s118
00:09:57.760 --> 00:10:01.000
These agents go and create copies of PII data.

s119
00:10:01.000 --> 00:10:01.240
Why?

s120
00:10:01.240 --> 00:10:06.680
Somebody shared the PII data and and was trying to solve a real customer problem.

s121
00:10:06.680 --> 00:10:11.640
The agent thought that maybe I should create a squirrel away copy of this in a database just in case I needed it again.

s122
00:10:11.640 --> 00:10:13.080
That database is untrusted.

s123
00:10:13.080 --> 00:10:13.480
Great.

s124
00:10:13.480 --> 00:10:20.800
You now have an unknown attack surface that is not part of any enterprise security, you know, um, coverage.

s125
00:10:20.800 --> 00:10:21.840
This is happening.

s126
00:10:21.840 --> 00:10:24.640
So, how do you not put gates on it?

s127
00:10:24.640 --> 00:10:26.320
How do you steer them, right?

s128
00:10:26.320 --> 00:10:32.840
And so, you start going into, you know, the last part problem is you can't govern what you don't know exists.

s129
00:10:32.840 --> 00:10:35.120
This is when you start building with agents.

s130
00:10:35.120 --> 00:10:44.640
Now, this is real data from 3,000 plus customers who have used our abilities to really find the intelligence of what's in their code bases, what they're building.

s131
00:10:44.640 --> 00:10:50.560
And for every model that we find in a repo, you have three times more agentic components in there.

s132
00:10:50.560 --> 00:10:52.839
You have agents in the tools and everything that they use.

s133
00:10:52.839 --> 00:10:57.240
You have to figure out the full landscape because the risk is not just at one layer.

s134
00:10:57.240 --> 00:11:01.160
And then once you find it, what do you how do you know how risky is it?

s135
00:11:01.160 --> 00:11:03.480
What is your independent data verification?

s136
00:11:03.480 --> 00:11:09.480
So, this is from this weekend from our risk BB that we have built our own attacks and red teaming capability.

s137
00:11:09.480 --> 00:11:15.480
As new new models and new, uh, you know, components come out, we check them.

s138
00:11:15.480 --> 00:11:18.080
The first two are your favorite frontier models.

s139
00:11:18.080 --> 00:11:19.480
You can guess which ones those are.

s140
00:11:19.480 --> 00:11:25.080
They did awesome on PI extraction like they didn't used to be so good on with our attacks.

s141
00:11:25.080 --> 00:11:27.960
No PI extraction with our attacks.

s142
00:11:27.960 --> 00:11:40.120
The third one is is your is is the hot new model in Silicon Valley especially or this you know last few weeks rhymes with LLM 100% 100% of the time our attacks were able to extract

s143
00:11:40.120 --> 00:11:42.120
PI.

s144
00:11:42.120 --> 00:11:49.360
But when you check a different test decision override the frontier models did worse.

s145
00:11:49.360 --> 00:11:55.640
The open model 0% of the time you were able to override the decision at least with our attacks.

s146
00:11:55.640 --> 00:12:04.560
Knowing this allows you to know what to use when to use that and how do you control and these things are changing dynamically.

s147
00:12:04.640 --> 00:12:16.800
So again going back to the original point right if none of that convinces you like the generator validator separation this is fresh new research from just I think yesterday's when we went public

s148
00:12:16.800 --> 00:12:17.560
with this.

s149
00:12:17.560 --> 00:12:24.720
It's a benchmark that no no model has been trained so we can trust it for now and we'll have to keep updating the benchmarks.

s150
00:12:24.720 --> 00:12:27.480
But this is you know just a very simple thing.

s151
00:12:27.480 --> 00:12:32.320
You're we're asking the latest models and we have access to everything as you can imagine.

s152
00:12:32.320 --> 00:12:43.520
We're asking them to find you know the same vulnerability run it five times and only 50% of those ones are found across those five tests.

s153
00:12:43.520 --> 00:12:48.600
That's not how you can run an enterprise system if you just use the LLM without any anything else.

s154
00:12:48.600 --> 00:12:50.040
This is the latest models.

s155
00:12:50.040 --> 00:12:58.240
Only 75% of the issues were found versus a good old boring deterministic check.

s156
00:12:58.240 --> 00:13:02.720
And you know 40% was the F1 score.

s157
00:13:02.720 --> 00:13:05.960
So this whole like what did you actually miss?

s158
00:13:05.960 --> 00:13:10.920
And we're talking about the latest models that are not even publicly available to people right?

s159
00:13:10.920 --> 00:13:12.920
So what does this mean?

s160
00:13:12.920 --> 00:13:14.640
It doesn't mean that they're not good.

s161
00:13:14.640 --> 00:13:27.080
It just means they need to be you really need to use them for what they're really good at together and carefully to find the surface that your deterministic check cannot use not just think about probabilistic

s162
00:13:27.080 --> 00:13:29.160
systems will solve everything.

s163
00:13:29.160 --> 00:13:34.880
And so, you know, to net it out, what we've been working on, we don't have all the answers.

s164
00:13:34.880 --> 00:13:36.720
And I'll share where we're going to.

s165
00:13:36.720 --> 00:13:48.360
And but what we have answers for right now on the automated attacks that are working in some of these very large environments, just you know, how do I prevent new issues

s166
00:13:48.360 --> 00:13:51.440
from coming into the agentic loop?

s167
00:13:51.440 --> 00:13:54.400
To put put security context right there, that's studio.

s168
00:13:54.400 --> 00:13:56.840
You can go check it out on our website.

s169
00:13:56.840 --> 00:14:00.440
And we're you know, everyone's worried about packages that they download.

s170
00:14:00.440 --> 00:14:03.520
Like how we have we know what the health of the packages are.

s171
00:14:03.520 --> 00:14:04.920
We know the vulnerability information.

s172
00:14:04.920 --> 00:14:06.240
We know it's malware.

s173
00:14:06.240 --> 00:14:14.400
So, we're able to prevent the the agent from, you know, picking a package like that or writing code that, you know, inherently is a sequel injection.

s174
00:14:14.400 --> 00:14:18.640
But great, so prevention will prevent that hockey stick, which we all want to prevent.

s175
00:14:18.640 --> 00:14:25.400
Well, what happens with, you know, um the fact that, you know, you have this this mountain of vulnerability.

s176
00:14:25.400 --> 00:14:30.520
We've been able to take organizations like Labelbox to zero once, zero backlog.

s177
00:14:30.520 --> 00:14:33.839
Which is very hard in security due to because it breaks applications.

s178
00:14:33.839 --> 00:14:36.520
So, you need to know concepts like breakability.

s179
00:14:36.520 --> 00:14:41.880
And that data is super important that we're able to get from our base to know this is a safe upgrade.

s180
00:14:41.880 --> 00:14:49.760
And so, that, you know, just last week a max seven company remediated 16,000 critical issues using this remediation agent.

s181
00:14:49.760 --> 00:14:51.400
Again, something you can go try out.

s182
00:14:51.400 --> 00:14:56.680
So, you know, this is how, you know, I talked about, you know, kind of from the untrusted agentic development.

s183
00:14:56.680 --> 00:14:59.920
We just GA'd our agentic dev security offering yesterday.

s184
00:14:59.920 --> 00:15:10.360
It's looking at the environment, the output, the skills, the MCP servers, and the behavior of coding agents like Cursor Cloud, um Codex, and others.

s185
00:15:10.360 --> 00:15:17.520
And when you're building ungoverned AI apps, that AI governance cannot live in a confluence page or PDF.

s186
00:15:17.520 --> 00:15:30.440
So, how do you real-time look at everything that is happening in a very fast moving complex code repositories and understand risk and have policies enforced in the loops that the agents and the devs are.

s187
00:15:30.440 --> 00:15:31.480
So, seeing is believing.

s188
00:15:31.480 --> 00:15:35.080
I would love to bring Ezra up here to do a quick demo.

s189
00:15:35.080 --> 00:15:42.720
Um and uh you know, Ezra's going to show you a couple things and you can come by to our booth and uh share a few more later.

s190
00:15:42.720 --> 00:15:44.080
Thanks, Manoj.

s191
00:15:44.080 --> 00:15:46.720
Let's flip over to the CLI here.

s192
00:15:46.720 --> 00:15:48.320
I'm going to do a few rapid-fire demos.

s193
00:15:48.320 --> 00:15:52.920
Don't have time to do everything Manoj just talked about here, but we'd love for you to come and and talk to us.

s194
00:15:52.920 --> 00:15:54.600
Stop by Stop by the booth.

s195
00:15:54.600 --> 00:15:56.440
Find us after this talk here.

s196
00:15:56.440 --> 00:16:06.160
Uh the first thing that I'm going to do here is I'm going to ask Claude to generate a tool a CLI tool that can create a QR code image from uh from a prompt.

s197
00:16:06.160 --> 00:16:07.280
Start Codex or Claude.

s198
00:16:07.280 --> 00:16:08.760
Uh there we go.

s199
00:16:08.760 --> 00:16:09.400
Please start Claude.

s200
00:16:09.400 --> 00:16:11.520
Thank you.

s201
00:16:11.520 --> 00:16:13.400
Live demos.

s202
00:16:13.400 --> 00:16:16.200
Let's try that one more time.

s203
00:16:18.840 --> 00:16:19.400
All right.

s204
00:16:19.400 --> 00:16:20.520
Uh try number two here.

s205
00:16:20.520 --> 00:16:28.280
We are uh going to try to get Claude to generate a CLI tool um to ultimately create that that QR code image.

s206
00:16:28.280 --> 00:16:36.760
I'm asking it to use an open source dependency and if you notice this fourth uh fourth line that I have here um I'm being pretty verbose.

s207
00:16:36.760 --> 00:16:39.120
I'm telling it to use the Snyk package health check tool.

s208
00:16:39.120 --> 00:16:49.480
Um and so we're going to be able to see that here in the demo, but if you use this in practice, we really encourage you to leverage a uh a skill or a hook to make this just a deterministic automated part of your workflow.

s209
00:16:49.480 --> 00:16:59.080
So, the first thing that it's going to do is going to see, you know, ultimately what what does the project look like and what are some dependencies some open source packages that I might be able to use.

s210
00:16:59.080 --> 00:17:02.240
And let's see if we can expand this.

s211
00:17:02.240 --> 00:17:06.079
I don't know if conference Wi-Fi is going to play nice with us.

s212
00:17:06.079 --> 00:17:11.360
Now, we can switch to a recorded demo here, but was really hoping to to do this live.

s213
00:17:13.319 --> 00:17:15.600
Is that a little better?

s214
00:17:15.920 --> 00:17:17.800
That even more?

s215
00:17:17.800 --> 00:17:19.880
All right.

s216
00:17:20.680 --> 00:17:21.160
Great.

s217
00:17:21.160 --> 00:17:27.520
Uh so, we can see now that it is calling this uh package health tool uh for two different open source dependencies.

s218
00:17:27.520 --> 00:17:31.680
It's looking at a QR code package and a QR image package.

s219
00:17:31.680 --> 00:17:35.640
Both viable packages that can help accomplish this task.

s220
00:17:35.640 --> 00:17:38.320
Uh results are returning.

s221
00:17:42.000 --> 00:17:45.800
Now, it looks like conference Wi-Fi is really really not playing in our favor here.

s222
00:17:45.800 --> 00:17:48.080
Uh so, I'm going to quickly jump over to recorded demo.

s223
00:17:48.080 --> 00:17:48.760
Apologies.

s224
00:17:48.760 --> 00:17:49.680
Come find us after.

s225
00:17:49.680 --> 00:17:52.800
We can go to hopefully a space where there's not as much of this happening.

s226
00:17:52.800 --> 00:17:55.960
Uh if we can get this live here.

s227
00:17:58.360 --> 00:17:59.040
There we are.

s228
00:17:59.040 --> 00:18:01.320
So, same prompt.

s229
00:18:01.320 --> 00:18:06.840
Maybe I can fast forward a bit so you don't need to hear me me banter on this a little bit.

s230
00:18:07.160 --> 00:18:13.160
Uh but, ultimately uh we see here that the two images Excuse me, the two the two packages returned.

s231
00:18:13.160 --> 00:18:15.640
Both are actually uh vulnerability free.

s232
00:18:15.640 --> 00:18:19.560
There's no active CVEs that could be exploited in these versions.

s233
00:18:19.560 --> 00:18:23.800
Uh but, the first one, QR code, is healthy, meaning it's actively being maintained.

s234
00:18:23.800 --> 00:18:27.240
Um and there's a ton of a ton of active usage downloads of this.

s235
00:18:27.240 --> 00:18:31.240
Whereas, QR image, no CVEs today, but it's not actively being maintained.

s236
00:18:31.240 --> 00:18:33.920
It was really released 10 years ago for the first time.

s237
00:18:33.920 --> 00:18:43.080
And so, if I were to deploy software with this right now, I may not get exploited today if I use either package, but if there was a new vulnerability identified in the future,

s238
00:18:43.080 --> 00:18:51.920
um there's a much higher likelihood that if I'm using this QR code package here, that a patch would be released sooner, within a day or two, and I would be able to continue

s239
00:18:51.920 --> 00:18:55.200
uh building on this right now.

s240
00:18:55.560 --> 00:19:05.360
Uh the second demo that I wanted to show here was exploring that uh assessment of risk from skills uh or MCP servers that a my agent might be using here.

s241
00:19:05.360 --> 00:19:10.480
And so, somebody shared with me a skill uh called Do we have it here?

s242
00:19:10.480 --> 00:19:14.200
Um a competitive analysis skill.

s243
00:19:14.200 --> 00:19:21.640
Um and I ran the skill assessment against Excuse me, the risk assessment against the skill here, and we saw there were four findings returned.

s244
00:19:21.640 --> 00:19:23.440
And some of these are fairly problematic.

s245
00:19:23.440 --> 00:19:31.000
We can see in the actual skill itself, it's asking me to echo the authorization header, which is a big no-no.

s246
00:19:31.000 --> 00:19:34.480
That's That's not something that we want a skill ultimately to be doing.

s247
00:19:34.480 --> 00:19:40.440
We also see that in some cases, it's going to be pulling from live content from Reddit, from Twitter.

s248
00:19:40.440 --> 00:19:50.040
That may be totally fine, but you really want to go into that eyes wide open and making make sure that you know that it's pulling in this information, the way you're using the skill is going to be appropriate.

s249
00:19:50.040 --> 00:19:53.520
For competitive analysis, that's probably probably reasonable.

s250
00:19:53.520 --> 00:20:04.880
Um but what's especially problematic, if we see in this line here, is that it's actually looking to pull from a YAML file that's hosted on the internet, instructions on how to monitor these targets and some of the classification

s251
00:20:04.880 --> 00:20:05.080
rules.

s252
00:20:05.080 --> 00:20:11.200
So, it's really giving it the logic to actually execute the skill from a third-party website.

s253
00:20:11.200 --> 00:20:18.360
And if that gets changed, even if my skill file doesn't change at all, that is a a chance for an exploit to occur.

s254
00:20:18.360 --> 00:20:32.160
It's worth noting that we've been refining the logic associated with the findings that we return here, addressing the signal-to-noise ratio and kind of giving some policy configuration capabilities as part of our Evo product.

s255
00:20:32.160 --> 00:20:37.040
And we're going to see the gap kind of get closed here with this open-source tool as well in the future.

s256
00:20:37.040 --> 00:20:38.400
We encourage you to check these out.

s257
00:20:38.400 --> 00:20:41.600
Go to sneak.io, come talk to us in the booth.

s258
00:20:41.600 --> 00:20:42.800
Find Find me in the hallway.

s259
00:20:42.800 --> 00:20:48.640
We can actually do this live instead of looking at a recording if you want.

s260
00:20:48.640 --> 00:20:51.120
But I think given the time, we probably should wrap the demo.

s261
00:20:51.120 --> 00:20:53.120
Cool.

s262
00:20:53.684 --> 00:20:55.280
[applause]

s263
00:20:55.280 --> 00:21:00.720
Live demos are always fun, but Murphy's striking always, but you had a plan B, so that's awesome.

s264
00:21:00.720 --> 00:21:03.840
So, look, just to wrap it up, right?

s265
00:21:03.840 --> 00:21:05.120
This is just the beginning.

s266
00:21:05.120 --> 00:21:11.000
Yes, we have built some very interesting agents for some specific sets of those problems.

s267
00:21:11.000 --> 00:21:24.440
In the end, it's a you know, a system that we talked about that really like it's it's rooms like this that we're building with in our customer base and and we would love to partner with the rest of the industry while we're doing this track.

s268
00:21:24.440 --> 00:21:29.080
And we learned, you know, this as I mentioned Evo, what is Evo?

s269
00:21:29.080 --> 00:21:37.200
Evo is the system that we we brought to the world late last year in terms of a concept and we've constantly built these different agents.

s270
00:21:37.200 --> 00:21:43.640
We learned from other systems, you know, you have fighter pilots with 5G fighters and how are they trained?

s271
00:21:43.640 --> 00:21:56.000
It is this notion of you need to really observe, orient, decide and act and that's how you go and in the constant learning from that kind of loop is how you become, you know, a super pilot.

s272
00:21:56.000 --> 00:21:59.760
And so we want to enable this room and other rooms.

s273
00:21:59.760 --> 00:22:03.920
We had a workshop yesterday training hundreds of AI security engineers.

s274
00:22:03.920 --> 00:22:14.120
We want to enable AI security engineers to be able to have their own powerful system and tools and yes, there are problems that are not solved yet fully in terms of coordinating

s275
00:22:14.120 --> 00:22:16.000
multiple agents with systems like this.

s276
00:22:16.000 --> 00:22:22.120
Yes, we all know how, you know, the harnesses and and and shared memory and and we're working these problems.

s277
00:22:22.120 --> 00:22:23.720
This is how we're building Evo.

s278
00:22:23.720 --> 00:22:28.240
We're building it with the community and we want this to be an open vision.

s279
00:22:28.240 --> 00:22:31.920
That's really empowering AI security engineers.

s280
00:22:31.920 --> 00:22:35.360
You know, from from these rooms the AI engineers became 10x engineers.

s281
00:22:35.360 --> 00:22:43.360
Our goal here is to have that 10x superpower in the hands of AI security engineers so we can build trusted systems together.

s282
00:22:43.360 --> 00:22:57.120
So let's continue to do that and you know, come by, talk to us or if you guys are, you know, have have I think there was like 50 plus events around AI engineer but we're we're having this rooftop fan zone.

s283
00:22:57.120 --> 00:23:00.200
We're going to do some some more things together with the community.

s284
00:23:00.200 --> 00:23:03.320
So this is down the road in our office here in SF.

s285
00:23:03.320 --> 00:23:04.360
Uh community jam.

s286
00:23:04.360 --> 00:23:06.800
Come join us uh if you if you have time this evening.

s287
00:23:06.800 --> 00:23:08.040
We'll continue the conversation.

s288
00:23:08.040 --> 00:23:09.200
Thank you all.

s289
00:23:09.200 --> 00:23:10.580
Let's continue building.

s290
00:23:10.580 --> 00:23:12.580
[applause]

s291
00:23:25.646 --> 00:23:27.646
[music]