1
00:00:00,240 --> 00:00:03,880
Welcome to Data Driven, the podcast that explores the collision of

2
00:00:03,880 --> 00:00:07,200
data, AI and occasionally common sense.

3
00:00:07,520 --> 00:00:11,120
Today's guest is Mike Armistead, CEO of Pulse Security

4
00:00:11,200 --> 00:00:14,680
AI, a man who's been defending digital fortresses since before

5
00:00:14,680 --> 00:00:18,480
AI was cool and hackers had LinkedIn profiles. We talk

6
00:00:18,480 --> 00:00:22,000
about AI as both weapon and watchdog, why LLMs need

7
00:00:22,000 --> 00:00:25,640
guardrails and possibly a muzzle, and how your next data breach

8
00:00:25,640 --> 00:00:29,350
might come gift wrapped in a prompt. Grab your headphones and your

9
00:00:29,350 --> 00:00:31,910
password manager and let's get Data Driven.

10
00:00:34,630 --> 00:00:38,270
Hello and welcome to Data Driven Podcast. We explore the

11
00:00:38,270 --> 00:00:41,910
emergent field of artificial intelligence, data engineering, and data

12
00:00:41,910 --> 00:00:45,750
science. And you'll notice that Andy looks a

13
00:00:45,750 --> 00:00:49,150
bit different today. If you're viewing the screen, if you're viewing this, if you're listening

14
00:00:49,150 --> 00:00:52,550
to this, he'll sound a bit different. That's because Andy is actually presenting

15
00:00:52,950 --> 00:00:56,620
a pre con today at SQL Past in Seattle. And,

16
00:00:56,770 --> 00:01:00,410
and I am in my car because many complicated

17
00:01:00,410 --> 00:01:03,970
reasons, but I'm not

18
00:01:03,970 --> 00:01:07,530
driving. And I have with me my co host on Impact,

19
00:01:07,530 --> 00:01:11,090
Quantum, which I believe that we'll all be in good

20
00:01:11,090 --> 00:01:14,930
hands here. How's it going, Candace? It's great. It's great. I'm

21
00:01:14,930 --> 00:01:18,730
actually really excited because with all honesty, although we focus so much

22
00:01:18,730 --> 00:01:22,530
on Quantum, the truth is AI and Quantum

23
00:01:22,530 --> 00:01:26,130
are now being like, spoken as if they're already

24
00:01:26,130 --> 00:01:29,890
one word. So being able to speak today to Mike,

25
00:01:30,050 --> 00:01:33,650
who I understand is the CEO of Pulse Security

26
00:01:33,810 --> 00:01:37,450
AI, makes. I'm very excited about the conversation, which. Is

27
00:01:37,450 --> 00:01:41,210
another field that is intricately tied to Quantum and AI.

28
00:01:41,210 --> 00:01:45,010
Right. This is like, this is the center of the Venn diagram. Right?

29
00:01:45,410 --> 00:01:49,010
So. So welcome to the show, Mike. Yeah, thank you. Thanks for having me.

30
00:01:49,490 --> 00:01:53,250
Hey, no problem, no problem. So just a quick question.

31
00:01:53,330 --> 00:01:55,410
What exactly does your company do?

32
00:01:57,250 --> 00:02:01,010
You know, so we are in. I'm one of these stealth companies

33
00:02:01,330 --> 00:02:05,090
still, so I will. But let me, let

34
00:02:05,090 --> 00:02:08,610
me generally describe to you the problem that we're

35
00:02:08,610 --> 00:02:12,330
addressing. And it's, it's a little bit. It's

36
00:02:12,330 --> 00:02:15,490
interesting because I think it, it definitely falls upon

37
00:02:17,070 --> 00:02:20,670
earlier waves of even what was going on in AI. My previous company

38
00:02:21,550 --> 00:02:24,670
was, was called response Software. We actually used

39
00:02:25,390 --> 00:02:29,150
AI back in 2016, which

40
00:02:29,150 --> 00:02:31,950
was a little bit different though. You know, the field of AI is very broad.

41
00:02:34,110 --> 00:02:37,830
We were probably more on the expert system end of that

42
00:02:37,830 --> 00:02:41,550
spectrum than what. Where the LLMs are today.

43
00:02:43,390 --> 00:02:46,730
And. But our journey was fantastic.

44
00:02:47,130 --> 00:02:50,930
We were applying AI to do something that, I'll say it in

45
00:02:50,930 --> 00:02:54,530
today's terms, everyone can understand, which is we were an

46
00:02:54,530 --> 00:02:58,170
assistant for a tier one soc Analyst,

47
00:02:58,170 --> 00:03:01,890
which if you know, in enterprises, security operations

48
00:03:01,890 --> 00:03:05,570
centers or a soc have really

49
00:03:05,570 --> 00:03:09,290
struggled to get skilled and even

50
00:03:09,290 --> 00:03:13,010
just people to be able to interpret what these signals

51
00:03:13,010 --> 00:03:16,590
are coming at them and what's a real threat and what's not a real

52
00:03:16,590 --> 00:03:20,350
threat and what's going on there. So they have, and they, most

53
00:03:20,350 --> 00:03:24,070
of them have to be 7 by 24. So it felt like

54
00:03:24,070 --> 00:03:27,910
a really great application where AI, because the AI can do a lot

55
00:03:27,910 --> 00:03:31,590
of that assistance and then give it to person, to a person to make the

56
00:03:31,590 --> 00:03:34,670
final judgment. And we learned a lot along the way. In fact,

57
00:03:35,230 --> 00:03:38,990
we ended up getting acquired by a company called

58
00:03:38,990 --> 00:03:42,590
Mandiant, which is already a public company in the

59
00:03:42,590 --> 00:03:46,230
security space. They're most known for doing

60
00:03:47,350 --> 00:03:51,110
incident responses. So that's when you know, someone gets hacked and

61
00:03:51,110 --> 00:03:53,910
they have to parachute in to try to get them back on their feet again,

62
00:03:54,550 --> 00:03:58,390
which is a very, you know, manual human kind of way of going

63
00:03:58,390 --> 00:04:02,150
through it. But they also had products and

64
00:04:02,870 --> 00:04:06,470
our team kind of got, got involved with that. That company

65
00:04:06,470 --> 00:04:10,310
ended up getting bought by Google a few years later.

66
00:04:10,950 --> 00:04:14,470
And so myself and our team was at Google for

67
00:04:14,870 --> 00:04:17,950
a couple of years. And we were at Google during an interesting time because that

68
00:04:17,950 --> 00:04:21,750
was code red happened at Google when we were there, which is

69
00:04:21,910 --> 00:04:25,630
chatgpt came out and Google had already had.

70
00:04:25,630 --> 00:04:28,390
Yeah, exactly. Google already had all this

71
00:04:29,270 --> 00:04:32,870
investment in AI that they weren't really telling anyone about.

72
00:04:33,510 --> 00:04:36,860
And ChatGPT beat him to the punch. And

73
00:04:36,860 --> 00:04:40,220
suddenly by edict of the CEO,

74
00:04:40,620 --> 00:04:43,900
every product had to have AI features in it.

75
00:04:44,380 --> 00:04:47,660
And our team was already in charge of the

76
00:04:48,060 --> 00:04:51,820
large language model for security. And so we got to see

77
00:04:51,820 --> 00:04:55,620
from all the teams that were doing product kind of what worked and

78
00:04:55,620 --> 00:04:58,860
what didn't. And there's a lot because,

79
00:04:59,830 --> 00:05:03,630
you know, I mean, it's like you guys, you see me, I've been in the

80
00:05:03,630 --> 00:05:07,190
industry for a long time, been through many waves, was an

81
00:05:07,190 --> 00:05:10,950
executive at a, at a

82
00:05:11,590 --> 00:05:14,950
Internet 1.0 company in the days of

83
00:05:15,190 --> 00:05:18,710
Web 1.0 and you know, ran

84
00:05:18,710 --> 00:05:22,390
ops and ad tech and all this stuff from that. So I understand these different

85
00:05:22,390 --> 00:05:25,830
waves, but LLMs aren't the answer to everything.

86
00:05:26,150 --> 00:05:29,960
And we got to see a lot of that. Makes

87
00:05:29,960 --> 00:05:33,680
you laugh. That's good, Frank. It should make you laugh. It's truer

88
00:05:33,680 --> 00:05:37,320
words that's ever been said. Right. So I remember when I first made the switch

89
00:05:37,320 --> 00:05:41,120
from Windows Phone development into AI or data science or machine

90
00:05:41,120 --> 00:05:44,840
learning, which was called then, it was a very different world.

91
00:05:44,840 --> 00:05:48,640
This is all pre LLM, right. I think

92
00:05:48,640 --> 00:05:52,360
there's going to be like an ad and a BC moment

93
00:05:52,520 --> 00:05:56,170
for AI people. It's probably going to be, you

94
00:05:56,170 --> 00:05:59,170
know, invention of ChatGPT or the release of Chat CPT.

95
00:05:59,810 --> 00:06:03,490
Right. And you know, where now everything's about LLMs. LLM

96
00:06:03,490 --> 00:06:06,890
that there's plenty other types of AI out there. Right. Whether it's good old fashioned

97
00:06:06,890 --> 00:06:10,690
math and stats, statistical analysis,

98
00:06:11,089 --> 00:06:14,770
it's actually easier to do than it is say and,

99
00:06:15,330 --> 00:06:19,050
or it's just, you know, old fashioned, you know, just machine learning. Right.

100
00:06:19,050 --> 00:06:21,900
They're not related to, you know,

101
00:06:22,140 --> 00:06:25,700
LLMs. Right. LLMs. I think it's kind of taking all the oxygen out of the

102
00:06:25,700 --> 00:06:29,460
room for good or for bad. But I remember like I was

103
00:06:29,460 --> 00:06:33,260
just, I was sitting at the Microsoft Research, a Microsoft Research

104
00:06:33,500 --> 00:06:37,300
conference because I worked at Microsoft at the time and now I'm at Red Hat

105
00:06:37,300 --> 00:06:40,900
and I'm only. They're not sponsoring this and they're not, you know,

106
00:06:40,900 --> 00:06:43,820
approving in this or this isn't completely independent. Just want to say that

107
00:06:44,860 --> 00:06:47,540
but my hair is a mess because haircut was one of the things I was

108
00:06:47,540 --> 00:06:50,510
supposed to do when my hot water tank decided to blow up flood my basin.

109
00:06:50,590 --> 00:06:54,430
Some entire weekend I've been putting stuff in dumpsters.

110
00:06:55,310 --> 00:06:59,110
But, but like you're right, like LLMs, you know, they're great tools,

111
00:06:59,110 --> 00:07:02,750
they're amazing. They're not going to solve everything. Right.

112
00:07:03,070 --> 00:07:06,830
And props to Google though. The, the paper that basically made

113
00:07:06,830 --> 00:07:10,550
LLMs, the technology on it was theirs. Attention is all you need. Was that

114
00:07:10,550 --> 00:07:14,190
2019ish. It was, it was a

115
00:07:14,190 --> 00:07:17,910
while ago. Yeah. You know, I mean, and a while

116
00:07:17,910 --> 00:07:21,550
ago maybe in today's terms. In today's terms, really it's like

117
00:07:21,550 --> 00:07:25,190
pre. Pandemic or post pandemic, honestly. Right, right. That's how people think

118
00:07:25,190 --> 00:07:28,950
about things. Right. How things, you know, and, and

119
00:07:28,950 --> 00:07:32,710
I. Think you know, why Google was holding on to things was there

120
00:07:32,710 --> 00:07:35,510
was a, there was a lot of unproven

121
00:07:36,070 --> 00:07:39,670
sides to using an LLM. And,

122
00:07:40,470 --> 00:07:44,250
and I think, you know, so in some ways as we look at, look forward

123
00:07:44,250 --> 00:07:47,850
and why there's so much thought about what's safe or what's not is

124
00:07:47,850 --> 00:07:51,250
because they were kind of holding onto it. Well,

125
00:07:52,050 --> 00:07:55,810
OpenAI didn't really feel they had that constraint and whether

126
00:07:55,810 --> 00:07:58,930
that's a good thing or a bad thing, we're going to find out. But they're

127
00:07:58,930 --> 00:08:02,530
both very different companies. Right. Google is a consumer enterprise company

128
00:08:03,330 --> 00:08:07,130
and OpenAI was just a research group of at the time, maybe

129
00:08:07,130 --> 00:08:10,550
what, 80 people, 100 people. Yeah. Right. Google was a

130
00:08:10,550 --> 00:08:14,190
worldwide phenomena. Like so if you, if you're that big, you really have to think

131
00:08:14,190 --> 00:08:17,750
very carefully before you release something like

132
00:08:17,750 --> 00:08:21,430
that. Whereas if you're just a research. Yeah, yeah, for sure.

133
00:08:22,230 --> 00:08:25,990
Anyway, to actually continue the story because that is the right

134
00:08:25,990 --> 00:08:29,670
sidecar. No, no, no, that's fine. Because I think it's an

135
00:08:29,670 --> 00:08:33,390
important kind of description of what's going on. We

136
00:08:33,390 --> 00:08:36,790
then eventually we formed this company called Pulse Security

137
00:08:36,870 --> 00:08:40,590
AI because we actually do believe that

138
00:08:40,590 --> 00:08:42,870
there's some really great applications

139
00:08:44,470 --> 00:08:47,990
of using an LLM. But I think within an agentix system

140
00:08:47,990 --> 00:08:51,030
rather than just the LLM being the database,

141
00:08:54,630 --> 00:08:58,230
we created a company that is into a place that there isn't

142
00:08:58,310 --> 00:09:01,670
been a lot of work, which is security

143
00:09:01,750 --> 00:09:05,510
programs are multi dimensional. There's a lot

144
00:09:05,510 --> 00:09:09,110
to them. They grew up kind of in a technology era

145
00:09:09,110 --> 00:09:12,830
where you solved almost one thing at a time. So if there's a

146
00:09:12,830 --> 00:09:16,510
threat of malware, you create something that sandbox the

147
00:09:16,510 --> 00:09:20,030
malware and detonates it and allows you to take care of it.

148
00:09:20,110 --> 00:09:23,910
If there's a threat to access or you're over privileging things, you

149
00:09:23,910 --> 00:09:27,670
have to think of your identity and access management. But it all kind

150
00:09:27,670 --> 00:09:30,550
of grew up from that. But there's a layer kind of missing which is how

151
00:09:30,550 --> 00:09:34,160
do you connect all all of this together? And security

152
00:09:34,320 --> 00:09:36,640
teams and in our experience

153
00:09:38,080 --> 00:09:41,840
they put people which are great on the judgment, they put people

154
00:09:41,840 --> 00:09:45,680
in there. And so there's a lot of manual tasks about connecting the dots between

155
00:09:45,680 --> 00:09:49,400
things. And we think AI can help a lot in at

156
00:09:49,400 --> 00:09:53,000
the program level how you know, what people

157
00:09:53,000 --> 00:09:56,560
should do from a strategy standpoint, not just from a

158
00:09:57,220 --> 00:10:00,260
detailed kind of technical detection standpoint.

159
00:10:01,380 --> 00:10:04,340
No, absolutely. My wife actually works in cybersecurity

160
00:10:05,220 --> 00:10:08,980
for the US government at at nist. So like some of these

161
00:10:08,980 --> 00:10:11,940
things I'm familiar with. So when you said soc, I was like, oh, I know

162
00:10:11,940 --> 00:10:15,660
what that is, you know, like. And mand I'm

163
00:10:15,660 --> 00:10:19,460
familiar with them, right. And it's an interesting,

164
00:10:20,100 --> 00:10:23,390
it's an interesting time because when I

165
00:10:24,030 --> 00:10:27,150
First, when ChatGPT released, I was just coming back from

166
00:10:27,870 --> 00:10:31,710
reinvent in Vegas and you know, anyone's been to Vegas, right?

167
00:10:31,710 --> 00:10:34,550
You know, like after the third day you get to the airport early because you

168
00:10:34,550 --> 00:10:38,350
just have to get out, you know what I mean? And

169
00:10:39,870 --> 00:10:42,510
I was starting to play with it and I was like, wow, I'm actually really

170
00:10:42,510 --> 00:10:45,110
impressed with it. So my wife picks up the airport. I'm. All I could talk

171
00:10:45,110 --> 00:10:48,590
about is chat gbd. Like that's literally all I could talk about. And she was

172
00:10:49,840 --> 00:10:53,200
so like, well, I'm like, it's trained on all this corpus of data. And, and

173
00:10:53,200 --> 00:10:55,840
she just looked at me and said, so that means all the data that it's

174
00:10:55,840 --> 00:10:58,480
trained on is basically one giant attack surface.

175
00:10:59,920 --> 00:11:03,520
And I was like, oh, my God, she's right.

176
00:11:03,680 --> 00:11:07,200
But when I would tell fellow data scientists and AI engineers that

177
00:11:07,360 --> 00:11:10,600
they would look like me, they would look at me like I had a tinfoil

178
00:11:10,600 --> 00:11:13,760
hat on. And I was like, you know, talking about

179
00:11:13,840 --> 00:11:16,880
conspiracies and lizard people. You know what I mean? Like, that's how they looked at

180
00:11:16,880 --> 00:11:20,720
me. But, you know, a few years later, right, what's

181
00:11:20,720 --> 00:11:24,480
on the owasp? It's like second or third, right?

182
00:11:24,880 --> 00:11:28,520
Yeah, right, for sure. I mean, there's. I often

183
00:11:28,520 --> 00:11:31,840
talk. Because you've talked about the security program,

184
00:11:32,560 --> 00:11:36,400
it ends up that you end up talking about

185
00:11:36,960 --> 00:11:40,760
strategy, and strategy has to include what your adversaries are doing plus

186
00:11:40,760 --> 00:11:44,570
what you have internally. And so I end

187
00:11:44,570 --> 00:11:47,890
up talking a bit about even

188
00:11:48,210 --> 00:11:51,650
use of AI by the adversary and, or

189
00:11:52,130 --> 00:11:55,650
leveraging the AI by the adversary. And so new

190
00:11:55,650 --> 00:11:59,330
kinds of attacks based on a

191
00:11:59,330 --> 00:12:02,930
prompt injection. I mean, that's a, that is a new thing where you

192
00:12:02,930 --> 00:12:06,650
could, you know, just through the prompt, ask it

193
00:12:06,650 --> 00:12:10,410
to divulge information. It shouldn't be divulging. But, but also you

194
00:12:10,410 --> 00:12:12,130
bring up a great point, Frank, which is just

195
00:12:14,080 --> 00:12:17,680
the LLMs, when they're getting trained, are using

196
00:12:18,000 --> 00:12:21,760
data and you have to be very sensitive

197
00:12:21,760 --> 00:12:25,600
to what data that is there. That's

198
00:12:25,600 --> 00:12:29,280
why I think a lot of enterprises are scrambling to make sure that their policies

199
00:12:30,960 --> 00:12:34,600
are set, that they can make all their employees aware of. Don't put sensitive

200
00:12:34,600 --> 00:12:38,450
information, even though it provide great context to your

201
00:12:38,530 --> 00:12:42,330
prompt, that it's going to be used and

202
00:12:42,330 --> 00:12:46,170
it's going to be, it's going to be sucked in there, and before you know

203
00:12:46,170 --> 00:12:49,810
it, it's going to be in everybody's, you know, prompt or

204
00:12:49,810 --> 00:12:53,650
available to everybody. And it's, it's definitely a real thing. So

205
00:12:54,450 --> 00:12:57,650
given your background in cybersecurity and talking about,

206
00:12:59,490 --> 00:13:03,130
you know, LLMs and the LLMs adoption, what is. Do you think

207
00:13:03,130 --> 00:13:05,970
that that is the biggest unaddressed security risk

208
00:13:07,070 --> 00:13:10,710
is not training the LLM properly so that

209
00:13:10,710 --> 00:13:13,510
it doesn't protect the data that it has. Like, what do you think is the

210
00:13:13,510 --> 00:13:15,310
biggest unaddressed security risk?

211
00:13:17,550 --> 00:13:19,950
I think a little bit related is

212
00:13:21,390 --> 00:13:25,110
ChatGPT and Gemini and Cloud. They've kind

213
00:13:25,110 --> 00:13:27,390
of. They're teaching everybody that

214
00:13:28,910 --> 00:13:32,430
their system is a database of answers,

215
00:13:32,750 --> 00:13:35,150
when, in fact, you shouldn't be thinking about it that way. You should think of

216
00:13:35,150 --> 00:13:38,800
it as it is a tool that helps you collect the answers and

217
00:13:38,800 --> 00:13:42,240
see the answers and do that. And

218
00:13:42,320 --> 00:13:45,200
so the real

219
00:13:45,520 --> 00:13:48,880
danger actually is in

220
00:13:50,080 --> 00:13:53,680
the fact that the adversaries can use the same

221
00:13:53,760 --> 00:13:57,560
technology to perform attacks at scale and speed

222
00:13:57,560 --> 00:14:00,720
that we haven't really been used to.

223
00:14:01,920 --> 00:14:05,740
And so there's that aspect to it. Then the

224
00:14:05,740 --> 00:14:08,660
other aspect is a data

225
00:14:09,300 --> 00:14:12,980
like hole that's there I think

226
00:14:12,980 --> 00:14:15,860
typical in the cybersecurity world.

227
00:14:16,820 --> 00:14:20,300
The business is really wanting to use this because it's such a

228
00:14:20,300 --> 00:14:23,540
productivity gain and whatever it might be, either

229
00:14:24,260 --> 00:14:27,900
your business side is either really pushing it for creative

230
00:14:27,900 --> 00:14:31,140
work or pushing it for just understanding

231
00:14:31,770 --> 00:14:35,450
different parts of the business. And they're ahead of the security team.

232
00:14:35,610 --> 00:14:39,450
And that happens quite frequently. You know, in the my, not the

233
00:14:39,450 --> 00:14:43,130
last company, but before that was all about application security. It was clear

234
00:14:43,450 --> 00:14:47,210
software developers, you know, they were pushing the envelope about

235
00:14:47,770 --> 00:14:51,410
making software so core to many organizations and they were thinking of building

236
00:14:51,410 --> 00:14:54,730
stuff. They weren't thinking of somebody using it

237
00:14:55,210 --> 00:14:59,050
to divulge, you know, corporate information

238
00:14:59,210 --> 00:15:02,570
or to take down a corporation, you know, for

239
00:15:02,940 --> 00:15:06,740
basically using it against them. They're creators, they don't think

240
00:15:06,740 --> 00:15:10,340
about destroyers and the adversaries are

241
00:15:10,340 --> 00:15:13,660
destroyers. And so you had to weave in security

242
00:15:13,900 --> 00:15:17,620
into that culture, which remains a challenge today. I think

243
00:15:17,620 --> 00:15:21,300
that's what's going on right now with LLMs. People are thinking, oh, I can use

244
00:15:21,300 --> 00:15:24,460
it for all these things. They aren't thinking what it's exposing

245
00:15:24,940 --> 00:15:28,780
Frank, back to your wife's thing. They're not thinking of the attack surface

246
00:15:28,780 --> 00:15:32,240
you're suddenly creating by doing that. And

247
00:15:32,480 --> 00:15:36,040
I think that's the biggest thing, Kansas, it's more that attack surface

248
00:15:36,040 --> 00:15:39,360
expansion or just having

249
00:15:39,760 --> 00:15:42,800
even the current attack surface just be more readily available

250
00:15:43,920 --> 00:15:47,440
to the attackers is the thing

251
00:15:47,840 --> 00:15:51,120
that's a real difference because ultimately it gets down to

252
00:15:52,720 --> 00:15:56,160
even these sophisticated attacks that you're starting to hear about now

253
00:15:56,690 --> 00:15:57,890
from the state sponsored

254
00:16:00,050 --> 00:16:02,210
entities that are out there.

255
00:16:03,730 --> 00:16:07,090
It's still coming down to they're exploiting age old vulnerabilities,

256
00:16:07,970 --> 00:16:11,730
but it's just that they're getting to them in a way that's more

257
00:16:11,810 --> 00:16:14,850
automatic. And

258
00:16:15,570 --> 00:16:19,370
they can, as we often say in security, the bad

259
00:16:19,370 --> 00:16:22,050
guys kind of have usually all the time in the world and they only have

260
00:16:22,050 --> 00:16:25,620
to be right once. Yeah, right, right. Well, it's

261
00:16:25,620 --> 00:16:29,260
interesting. You're thinking about like the jewelers, the builders, the

262
00:16:29,260 --> 00:16:32,900
developers. That's their mindset versus the jewel heist

263
00:16:33,220 --> 00:16:37,060
people. Right? And that's two very different mindsets.

264
00:16:37,700 --> 00:16:41,380
And you know, I always joke that

265
00:16:41,460 --> 00:16:45,220
our kids are going to be like the first developers ever to write secure code.

266
00:16:45,220 --> 00:16:48,420
Right. That's my background.

267
00:16:49,220 --> 00:16:52,810
I was a developer. But in all

268
00:16:52,810 --> 00:16:56,170
seriousness. But one of the things that I heard is one of the things that's

269
00:16:56,170 --> 00:16:59,890
driving companies, because you mentioned companies or businesses are encouraging

270
00:17:02,930 --> 00:17:06,290
business users to use AI. One of the reasons I heard was

271
00:17:07,010 --> 00:17:10,690
there was so much shadow it going on, I'm sure it's still going on

272
00:17:11,090 --> 00:17:14,890
that if they banned it outright the stuff would just end up in

273
00:17:14,890 --> 00:17:18,630
a public form of ChatGPT or Gemini

274
00:17:18,790 --> 00:17:22,630
or Claude or something like that versus if they do it through the company way.

275
00:17:23,510 --> 00:17:27,190
The companies that purvey these models, the enterprise versions,

276
00:17:27,270 --> 00:17:30,910
they promise and pinky swear that they'll never use that

277
00:17:30,910 --> 00:17:34,670
data for training data set in the future. So I guess that's

278
00:17:34,670 --> 00:17:38,310
kind of better. But you're right, as I think about this,

279
00:17:38,390 --> 00:17:42,190
we're putting AI in all of these places and we're not really

280
00:17:42,190 --> 00:17:45,880
even sure exactly how it works. And even crazier still,

281
00:17:46,120 --> 00:17:49,680
we're not even. Sure. We'Re not

282
00:17:49,680 --> 00:17:52,880
even sure we know what

283
00:17:52,880 --> 00:17:56,280
vulnerabilities are currently out there. So we're not even sure

284
00:17:56,680 --> 00:18:00,520
now we're just pouring like all these new vulnerabilities in there.

285
00:18:00,680 --> 00:18:04,400
We don't know what we don't know obviously. And it's just kind of like, it's

286
00:18:04,400 --> 00:18:08,160
kind of wild like that. Yeah, I think it's also wild

287
00:18:08,160 --> 00:18:11,870
because the AI LLMs

288
00:18:11,870 --> 00:18:14,750
as trained, they speak so

289
00:18:14,750 --> 00:18:18,110
authoritatively and in such, you know, proper

290
00:18:18,190 --> 00:18:21,630
English and that you're, you're just apt to believe them.

291
00:18:21,950 --> 00:18:25,150
You know, I, I, I, you know, one of my

292
00:18:25,150 --> 00:18:28,870
soapbox, I guess I'll say it is that I think

293
00:18:28,870 --> 00:18:32,670
the, one of the biggest things we can do in society today is we've

294
00:18:32,670 --> 00:18:36,260
got to be teaching our kids at the junior

295
00:18:36,260 --> 00:18:39,980
high, high school levels for sure and certainly in college. It should

296
00:18:39,980 --> 00:18:43,140
be happening. But to be critical thinkers

297
00:18:43,860 --> 00:18:47,500
because you can't, you know, if the world of social

298
00:18:47,500 --> 00:18:51,340
media taught us anything, you know, people kind

299
00:18:51,340 --> 00:18:54,820
of believe stuff that maybe they shouldn't believe. And,

300
00:18:55,060 --> 00:18:58,860
and now you have an AI generating this. That sounds so

301
00:18:58,860 --> 00:19:02,500
believable. And heck, these days, you know, you,

302
00:19:02,860 --> 00:19:06,580
you even might see an image and think it's that person saying it.

303
00:19:06,580 --> 00:19:10,420
It might not be at all. And yet, yet you believe that. You

304
00:19:10,420 --> 00:19:13,500
got to, you got to kind of, you have to be, you have to

305
00:19:13,980 --> 00:19:17,340
maybe you trust, but you got to verify. You know, it's an age old thing.

306
00:19:17,420 --> 00:19:20,980
You just, you just can't believe things for their first blush.

307
00:19:20,980 --> 00:19:23,980
And, and yeah, it's a whole believe.

308
00:19:24,620 --> 00:19:28,100
H none of what you hear and only half of what you see. I think

309
00:19:28,100 --> 00:19:31,550
now it's, you have to believe none of what you see or hear. Right.

310
00:19:31,870 --> 00:19:35,510
Unless it happens physically in front of you. And

311
00:19:35,510 --> 00:19:39,310
even then. Yeah, I mean look, what, what many

312
00:19:39,310 --> 00:19:42,790
of the, the banks and other people that have to really have

313
00:19:42,790 --> 00:19:46,270
trusted systems are doing is, you know, they're,

314
00:19:46,430 --> 00:19:50,230
they're requiring on say a wire transfer. I know I

315
00:19:50,230 --> 00:19:54,070
just had to do this is they, they want to call, they want me to

316
00:19:54,070 --> 00:19:57,680
hold my, my, my license up next to my face.

317
00:19:59,520 --> 00:20:03,040
And even then, you know, there's techniques that you use and we can get back

318
00:20:03,040 --> 00:20:05,280
to the LLMs because you use a lot of. Well, I heard that some of

319
00:20:05,280 --> 00:20:07,920
them will make you do this now. Yeah. Or

320
00:20:08,880 --> 00:20:12,640
ask a question that is so off topic

321
00:20:13,280 --> 00:20:17,000
like, and just see if, what

322
00:20:17,000 --> 00:20:20,800
the response is, if it can't even respond, you know, ask for the favorite

323
00:20:20,800 --> 00:20:24,630
football, pro football team or something like that, you know, and, and

324
00:20:24,630 --> 00:20:27,630
just, you're going to be able to tell

325
00:20:29,070 --> 00:20:32,670
using that. And if you go. So even going back. So we

326
00:20:32,670 --> 00:20:34,270
use LLMs in our system

327
00:20:36,350 --> 00:20:40,030
and we, we, I think the next

328
00:20:40,030 --> 00:20:43,430
wave of things we really believe are those

329
00:20:43,430 --> 00:20:47,150
guardrails that you have to put on it so that it won't hallucinate.

330
00:20:47,980 --> 00:20:51,340
And you know, people think, oh, the hallucination, that's, that's an edge case.

331
00:20:51,580 --> 00:20:55,380
It is not. You know, they, they weren't

332
00:20:55,380 --> 00:20:58,460
really always hallucinating. I mean technically they were always hallucinating.

333
00:20:59,660 --> 00:21:03,339
I guess you could, you could say that. I mean it's, it's a

334
00:21:03,339 --> 00:21:07,020
probabilistic kind of way of, you know, getting the pattern and things.

335
00:21:07,020 --> 00:21:10,700
But, but what, but what it does, it's been, the

336
00:21:10,860 --> 00:21:14,700
models, the, the weights have been put on giving

337
00:21:14,700 --> 00:21:18,300
a good response or a response that fulfills the

338
00:21:18,300 --> 00:21:22,140
request and that waiting forces it

339
00:21:22,220 --> 00:21:24,300
to make up stuff when it doesn't know.

340
00:21:26,460 --> 00:21:30,220
And yet it sounds authoritative and things like that. And

341
00:21:30,220 --> 00:21:33,940
so you really have to have the guardrails on it. And so I think as

342
00:21:33,940 --> 00:21:37,140
I was saying, the next wave of systems are going to be very vertically aligned

343
00:21:37,140 --> 00:21:40,940
like us in cybersecurity. It might be health care, it might be other things,

344
00:21:41,280 --> 00:21:44,560
but they're going to know to make the LLM

345
00:21:47,120 --> 00:21:50,800
to ask it basically tell me when you're like, we have, we

346
00:21:50,800 --> 00:21:54,640
call them verification prompts, right. Or context. And so

347
00:21:54,640 --> 00:21:57,680
it requires them to say if you're making it up, you got to tell me

348
00:21:57,920 --> 00:22:01,680
basically, right. And then even then

349
00:22:03,040 --> 00:22:06,840
limit what it's using as its

350
00:22:06,840 --> 00:22:10,580
context because that'll help too. Because you can do that and

351
00:22:10,580 --> 00:22:13,740
make it more authoritative sources rather than

352
00:22:15,340 --> 00:22:18,940
on some Reddit board or something like that

353
00:22:19,420 --> 00:22:21,900
where it's clearly gathering information from.

354
00:22:23,100 --> 00:22:26,780
You have to do that. And there's people that do that. You see some of

355
00:22:26,780 --> 00:22:29,020
the AIs being very good about

356
00:22:31,260 --> 00:22:35,040
noting or citing its sources. I think that's something. I really

357
00:22:35,040 --> 00:22:37,640
like it when it does that. Yeah, totally. Right, because.

358
00:22:38,440 --> 00:22:42,280
Yeah. And they let you decide on the judgment because in my view,

359
00:22:42,280 --> 00:22:45,560
people have to be in the middle of this for a long time.

360
00:22:45,960 --> 00:22:49,160
Right. I'm not a believer. It's going to go sentient here

361
00:22:50,120 --> 00:22:53,720
shortly. Again, it's my web 1.0 side to me that

362
00:22:54,680 --> 00:22:57,880
the world was going to change. There were going to be no retailers, no bricks

363
00:22:57,880 --> 00:23:01,080
and mortar. If you guys remember that term, bricks and mortar retailers.

364
00:23:01,640 --> 00:23:05,040
You know, that was then they had clicking mortars. I was at barnes and

365
00:23:05,040 --> 00:23:08,440
noble.com during, during that era. Yeah. So you know this

366
00:23:08,920 --> 00:23:12,640
clicking water. Yeah. You know, and. But they,

367
00:23:12,640 --> 00:23:16,280
you know, the hype was they were going to get. That was just going to

368
00:23:16,280 --> 00:23:19,560
go by the wayside. And it was 10 years later, before that

369
00:23:19,880 --> 00:23:23,600
really start that Amazon stopped becoming a bookstore

370
00:23:23,600 --> 00:23:27,400
and started becoming, you know, much more than that or, or ebay got around.

371
00:23:27,880 --> 00:23:31,640
It was much, much later. Same thing happening in AI. It's not going to.

372
00:23:31,880 --> 00:23:35,000
These things aren't going to get there right away. So there's going to be vertical

373
00:23:36,440 --> 00:23:39,480
use of the AI that's going to

374
00:23:40,280 --> 00:23:44,119
provide the guardrails, provide the context that's necessary. And then people

375
00:23:44,120 --> 00:23:46,120
start trusting those kinds of things.

376
00:23:48,200 --> 00:23:51,240
And I think that's going to be needed for a while and then we're going

377
00:23:51,240 --> 00:23:54,760
to see a rise of something that people then can start to trust. But

378
00:23:55,320 --> 00:23:59,040
the LLM is not all that trustworthy right now and you need a lot

379
00:23:59,040 --> 00:24:02,840
of stuff around it to make it accurate and

380
00:24:03,640 --> 00:24:07,320
you know, not making up stuff. I'm

381
00:24:07,320 --> 00:24:10,680
sorry, do you believe the future LLMs will develop

382
00:24:10,919 --> 00:24:14,120
stronger reasoning capabilities or do you think that,

383
00:24:15,000 --> 00:24:18,760
you know, we'll still need the human critical thinkers always,

384
00:24:19,000 --> 00:24:22,660
you know, to close the loop. I

385
00:24:22,660 --> 00:24:26,260
think the ultimate. We're always going to need the human

386
00:24:26,660 --> 00:24:30,420
on judgment. So, you know, I think you can

387
00:24:30,420 --> 00:24:34,060
close certain loops pretty accurately even

388
00:24:34,060 --> 00:24:37,780
today with the LLM. But, but is it judgment

389
00:24:37,780 --> 00:24:41,380
and it's, you know, the LLMs are, you know, they're just repeating patterns and,

390
00:24:41,380 --> 00:24:45,140
and with what they have and, and things like that. So in

391
00:24:45,140 --> 00:24:47,960
fact I just did a, you know, did a prompt recently

392
00:24:49,000 --> 00:24:52,760
that I was asking one LLM to use another LLM

393
00:24:53,480 --> 00:24:57,120
and it came back with kind of an odd response. So it's

394
00:24:57,120 --> 00:25:00,840
like, so re asked it like, what version are you using? And

395
00:25:00,840 --> 00:25:04,400
sure enough, it was using a version that was like three

396
00:25:04,400 --> 00:25:07,880
versions ago. Because what it got trained on and

397
00:25:08,040 --> 00:25:11,840
you just make these assumptions. It's like, oh, of course we're now

398
00:25:11,840 --> 00:25:15,470
at ChatGPT 5. But something might not have been

399
00:25:15,470 --> 00:25:18,910
trained on that. It might have been trained on an old version. And so there's

400
00:25:18,910 --> 00:25:22,470
even that kind of thing happening. Sorry, Candice.

401
00:25:22,710 --> 00:25:26,470
To fully answer your question, though, I do believe that

402
00:25:26,950 --> 00:25:30,750
we are in for some things you might be able to close a

403
00:25:30,750 --> 00:25:33,270
loop for. But if they involve judgment,

404
00:25:34,470 --> 00:25:38,310
we almost ethically need to have a person involved

405
00:25:38,310 --> 00:25:42,140
with that because you just don't know where it's going to go. And, and, and

406
00:25:42,140 --> 00:25:45,740
you can't. And because they speak so well, people

407
00:25:45,740 --> 00:25:49,300
are already misunderstanding that,

408
00:25:49,380 --> 00:25:53,100
that, that, you know, they are like, like what, what they are really.

409
00:25:53,100 --> 00:25:56,900
And they're just repeating stuff that they know. Right. They're not, they're, they're kind of

410
00:25:56,900 --> 00:26:00,740
not making judgment calls. And there's so many things that are just

411
00:26:00,740 --> 00:26:04,540
about judgment that I think it's just better to think of

412
00:26:04,540 --> 00:26:08,390
them as a tool, not as this thing. I,

413
00:26:08,390 --> 00:26:12,190
I think there's a lot to get to, to get to these, you know,

414
00:26:12,190 --> 00:26:15,790
I know, I don't know. Sam Altman might say it's only two years

415
00:26:15,790 --> 00:26:18,750
away. I just think that's, that's, there's no way

416
00:26:19,550 --> 00:26:23,150
not, not for proper ethical judgment.

417
00:26:23,389 --> 00:26:26,750
Right? I mean, yeah, it might fake it

418
00:26:27,230 --> 00:26:30,830
really well, but it won't be ethics

419
00:26:30,830 --> 00:26:34,550
based judgment. And so do you think we

420
00:26:34,550 --> 00:26:37,790
could use AI tools to design better prompts.

421
00:26:39,070 --> 00:26:42,590
That we do that all the time? Absolutely

422
00:26:42,830 --> 00:26:46,630
you can. And in fact, I think it's

423
00:26:46,630 --> 00:26:49,710
almost the best practice now that you are both,

424
00:26:50,910 --> 00:26:54,550
like I had mentioned before, kind of the truth or the

425
00:26:54,550 --> 00:26:58,390
truth directive that you give it, you can give it a lot of pros. We

426
00:26:58,390 --> 00:27:02,140
also notice it's kind of, I don't know, like

427
00:27:02,140 --> 00:27:05,860
what about a month ago there

428
00:27:05,860 --> 00:27:09,420
was a very illustrative

429
00:27:09,900 --> 00:27:13,500
way of you need to threaten these things because it'll access

430
00:27:13,900 --> 00:27:17,620
or raise the stakes for these things because it'll access different parts of the

431
00:27:17,620 --> 00:27:21,220
model. Back to your thing, Frank. We don't really understand how they really

432
00:27:21,220 --> 00:27:24,900
work. And so it was just mind blowing in a way

433
00:27:24,900 --> 00:27:27,790
that you have to say. And so we even

434
00:27:28,670 --> 00:27:32,270
give our prompts the ability to say, hey,

435
00:27:32,430 --> 00:27:36,230
I will lose my job if I don't get this right. So get

436
00:27:36,230 --> 00:27:39,910
this right. But we definitely play the models

437
00:27:39,910 --> 00:27:43,710
off on each other because it's good

438
00:27:44,030 --> 00:27:47,750
and it's kind of asking one to be

439
00:27:47,750 --> 00:27:51,390
the devil's advocate on the other. And that's a known

440
00:27:51,390 --> 00:27:55,180
group think, you know, think about just people socially. Right

441
00:27:55,180 --> 00:27:58,540
group thinks, been around forever. And the way you,

442
00:27:58,780 --> 00:28:02,060
you go against it is you ask someone to be the

443
00:28:02,300 --> 00:28:06,020
devil's advocate in whatever this judgment needs to be. And

444
00:28:06,020 --> 00:28:09,860
that's a great way to test, pressure test if what you're hearing

445
00:28:09,860 --> 00:28:13,460
is actually right or not. And so yes, we have to pressure

446
00:28:13,460 --> 00:28:17,260
test, use the elements to pressure test each other, use our own

447
00:28:17,260 --> 00:28:20,940
prompts to pressure test the current model. You know, there's lots of different, different

448
00:28:20,940 --> 00:28:24,620
techniques to do this. I mean, you know, I think of your

449
00:28:24,620 --> 00:28:28,420
world especially, you guys have long specialized in, you

450
00:28:28,420 --> 00:28:31,940
know, data science has been one of these areas that uses a lot of these

451
00:28:31,940 --> 00:28:35,740
techniques to make sure that it just, you know, you don't get too narrow

452
00:28:35,740 --> 00:28:38,780
in the focus and you know, and you get to get the right answers.

453
00:28:39,180 --> 00:28:42,900
There's a whole, there's a whole new set of things that have to be

454
00:28:42,900 --> 00:28:46,660
done to make sure that we're, we're, we're using the tool in the way

455
00:28:46,660 --> 00:28:47,500
we should use it.

456
00:28:51,740 --> 00:28:55,580
I love you guys. Speechless. It's

457
00:28:55,580 --> 00:28:59,420
interesting. Like, and what's your take on private AI? Right, like running

458
00:28:59,420 --> 00:29:03,020
your AIs entirely on prem within servers you can control.

459
00:29:03,580 --> 00:29:07,420
I mean, I know a lot of people, including myself, think that's

460
00:29:07,420 --> 00:29:11,260
the cure all for a lot of these issues, but even then I'm thinking like,

461
00:29:11,500 --> 00:29:15,070
if it sounds like a cure all or a silver bullet, it's probably not.

462
00:29:16,190 --> 00:29:20,030
Yeah, I mean, I think it, it solves a bunch of these problems that we've

463
00:29:20,030 --> 00:29:23,790
been talking about. You know, it clearly does, but you

464
00:29:23,790 --> 00:29:27,630
can't air gap it totally because people, you want people to be using it.

465
00:29:28,030 --> 00:29:31,790
And so it'll, you know, you're still going to have insider threats.

466
00:29:32,270 --> 00:29:36,070
And so if you have an insider, you know, there's still going

467
00:29:36,070 --> 00:29:39,750
to be ways of getting information out. And it might not be

468
00:29:39,750 --> 00:29:43,230
a risk that you want to take as a company. I mean you still,

469
00:29:43,470 --> 00:29:46,430
and, and so you still have certain things

470
00:29:47,390 --> 00:29:51,030
that do it. But I do think it solves some things. The thing it doesn't

471
00:29:51,030 --> 00:29:54,590
solve is the, you know, why we're seeing

472
00:29:54,670 --> 00:29:57,950
such a rapid advancement in stuff is because

473
00:29:58,670 --> 00:30:02,470
it's the LLMs are looking at everything kind of that

474
00:30:02,470 --> 00:30:06,070
are public, that's public out there and making use of

475
00:30:06,070 --> 00:30:09,750
those and then people are looking at them and going, oh wow, that's great. And

476
00:30:09,750 --> 00:30:13,230
doing that, you'd have to replicate a bit of that. And yeah, you could bring

477
00:30:13,230 --> 00:30:16,490
those in and, and but there's going to be a lot of advances that we

478
00:30:16,490 --> 00:30:20,330
can't even predict right now. You know, like talking to, you know, Candace, you

479
00:30:20,330 --> 00:30:23,930
on the quantum side or you know, now we're seeing that, you know,

480
00:30:23,930 --> 00:30:27,770
Nvidia's got the chips right now, but the wafers and

481
00:30:27,770 --> 00:30:31,530
the amount of transist, you know, transistor equivalents you can put

482
00:30:31,530 --> 00:30:35,290
on these things are, it's going to impact things and maybe it's

483
00:30:35,290 --> 00:30:38,810
going to be practical, you know. No, nobody thought we'd have a whole

484
00:30:38,810 --> 00:30:42,410
computer on our phones, but you know, us going back

485
00:30:42,410 --> 00:30:46,250
into the 80s, yeah, it was a, that's a

486
00:30:46,250 --> 00:30:50,010
pretty powerful computer compared to what we were using at the time. You

487
00:30:50,010 --> 00:30:52,770
know, there's a lot of those things that are going to come to play. And,

488
00:30:52,930 --> 00:30:56,530
and so I, I do think bringing some of this stuff internal

489
00:30:57,010 --> 00:31:00,690
and that it'll solve some things. It won't solve everything though.

490
00:31:01,410 --> 00:31:04,530
And you know, you'll still have to, you'll still have to do a lot of

491
00:31:04,530 --> 00:31:07,970
good security hygiene. You'll start to do a good, a lot of good data hygiene.

492
00:31:08,620 --> 00:31:12,100
You know, I mean I'm kind of. Worrying though because like companies have not been

493
00:31:12,100 --> 00:31:15,100
doing a really bang up job of that last 50 years.

494
00:31:16,540 --> 00:31:20,180
Yeah, it's more noticeable, it's more noticeable now more than ever.

495
00:31:20,180 --> 00:31:23,580
I wonder what new vulnerabilities would private

496
00:31:23,580 --> 00:31:27,340
AI, what new vulnerable would private

497
00:31:27,340 --> 00:31:30,900
AI solve? And what or what, what new

498
00:31:30,900 --> 00:31:34,660
vulnerabilities would it, would it expose? Right? Like because we

499
00:31:34,660 --> 00:31:38,270
still don't know even if it's running on your server, you still don't know how

500
00:31:38,270 --> 00:31:41,150
it works. You know the only thing. And

501
00:31:42,670 --> 00:31:45,470
you're right and you also, that's why I brought up the

502
00:31:45,790 --> 00:31:49,630
insider. You know, it's an attack surface. You

503
00:31:49,630 --> 00:31:53,350
know, maybe you closed it down a little bit from being external but you have

504
00:31:53,350 --> 00:31:57,190
insider threat threats. You have other right. You know, the creative

505
00:31:57,190 --> 00:31:59,710
things that are going on on the attacker side about

506
00:32:01,470 --> 00:32:05,080
they've long kind of done, you know, the,

507
00:32:05,080 --> 00:32:08,880
where the attacks were. They'll get inside and they'll just wait and

508
00:32:08,880 --> 00:32:12,280
they'll wait for kind of the dust to clear so you cannot trace it back.

509
00:32:12,280 --> 00:32:15,880
And they'll cover their tracks and, and it

510
00:32:15,960 --> 00:32:18,920
could be sitting there in the first time someone in the business

511
00:32:19,880 --> 00:32:23,560
connects that model that's you think is walled

512
00:32:23,560 --> 00:32:26,920
off to something even for good

513
00:32:26,920 --> 00:32:30,470
legitimate business reasons. It might expose

514
00:32:31,830 --> 00:32:35,590
an avenue that someone could get in and start exfil trading. And you may not

515
00:32:35,590 --> 00:32:38,310
even know they are, I mean these low and slow

516
00:32:38,710 --> 00:32:42,270
Attacks that have been the bane of so many

517
00:32:42,270 --> 00:32:45,990
enterprises where you're just siphoning it off enough so

518
00:32:45,990 --> 00:32:49,350
that the controls don't see it. Those will happen

519
00:32:49,670 --> 00:32:52,790
in a lot of. And they could happen to models. And there you have your

520
00:32:52,790 --> 00:32:56,330
crown jewels, your data. That's everything slow slowly being

521
00:32:56,330 --> 00:32:59,970
siphoned off. You know, that's, that's going to remain. And

522
00:32:59,970 --> 00:33:03,530
you're going to have to have a multi layer security

523
00:33:04,490 --> 00:33:08,010
system in place to kind of deal with that as well.

524
00:33:09,450 --> 00:33:11,690
No, that's true. And it makes me wonder like,

525
00:33:14,330 --> 00:33:18,130
you know, I guess, I guess you can

526
00:33:18,130 --> 00:33:21,930
be. I had an actually interesting conversation with the customer a couple

527
00:33:21,930 --> 00:33:24,750
years ago and he talked about what's called the. And I know I'm going to

528
00:33:24,750 --> 00:33:28,510
mess up what the acronym is, but it's CIA Triad. And there's

529
00:33:28,510 --> 00:33:31,310
nothing to do with the Central Intelligence Agency. It's

530
00:33:32,030 --> 00:33:35,870
confidentiality, something.

531
00:33:36,350 --> 00:33:40,189
And what is it? Probably identity. Yeah.

532
00:33:40,750 --> 00:33:44,030
Or integrity, I think. And then access.

533
00:33:45,470 --> 00:33:49,230
Right. And he had this whole, you know, he had a whole thing where like,

534
00:33:49,230 --> 00:33:52,860
you know, if you lock things down so much, you basically kill

535
00:33:52,860 --> 00:33:56,620
the access part of it. Right. You basically make it impossible to access. Right. If

536
00:33:56,620 --> 00:34:00,420
you. It seems like security is one of those jobs that

537
00:34:01,140 --> 00:34:04,860
will be augmented by AI for sure. Right. Because no one's going to have time

538
00:34:04,860 --> 00:34:07,300
to read gigs and gigs of log files anymore. Right.

539
00:34:09,300 --> 00:34:12,940
But it's also going to need. You're going to need a human in the loop.

540
00:34:12,940 --> 00:34:14,980
Right. I don't say that because that's what my wife does. And

541
00:34:16,580 --> 00:34:17,060
I like.

542
00:34:20,180 --> 00:34:23,540
Yeah, I like paying. You bring up a great point.

543
00:34:23,860 --> 00:34:26,260
And let me transition it to this because

544
00:34:27,700 --> 00:34:31,420
I think I'm going to use a term that gets misapplied

545
00:34:31,420 --> 00:34:34,340
a lot for enterprises and it's about risk.

546
00:34:35,060 --> 00:34:38,820
You are not going to. And Candice, it gets to your

547
00:34:38,820 --> 00:34:42,100
point too. The job of security

548
00:34:43,220 --> 00:34:46,740
programs inside of enterprise is actually to mitigate

549
00:34:49,070 --> 00:34:52,870
the risks to the business. It's not to provide 100%

550
00:34:52,870 --> 00:34:56,510
security. That's not the goal. The goal is to mitigate the risk because

551
00:34:56,510 --> 00:35:00,310
every business is going to have risk. And, and you need to accept a certain

552
00:35:00,310 --> 00:35:04,070
amount of risk so that you can do business and you can reach more

553
00:35:04,070 --> 00:35:06,350
people and you can, you can do that. And

554
00:35:07,550 --> 00:35:11,150
circling all the way back to what Pulse Security

555
00:35:11,310 --> 00:35:15,070
does is that we hope to bring that concept back into things, is

556
00:35:15,070 --> 00:35:18,120
that the leaders should be thinking about risk

557
00:35:18,600 --> 00:35:22,240
and tracking their risks and knowing where they're

558
00:35:22,240 --> 00:35:25,400
taking risks or where they're not taking risk. I think today

559
00:35:25,960 --> 00:35:29,720
why I said it's kind of one of these misplaced things is we kind

560
00:35:29,720 --> 00:35:33,080
of allow regulations and things like that

561
00:35:33,640 --> 00:35:37,360
to say to be about risk. And that's really the low

562
00:35:37,360 --> 00:35:41,040
bar, you know, in security we always talk about, you know, if

563
00:35:41,040 --> 00:35:44,630
you're, you talked about the OWASP earlier or you know, you think about the

564
00:35:44,630 --> 00:35:48,070
PCI standard, you know, for retailers and

565
00:35:48,150 --> 00:35:51,350
transaction processing, or you think about some of these other

566
00:35:51,670 --> 00:35:55,190
standards, they're the low bar. And many people

567
00:35:55,270 --> 00:35:58,310
think about risk as something that I have to do it we call

568
00:35:58,390 --> 00:36:02,230
checkbox compliance. Right. I have to be compliant, but I only want to

569
00:36:02,230 --> 00:36:04,990
do as much as I, as I can. As you need to. Because no one

570
00:36:04,990 --> 00:36:08,680
looks forward to seeing security people, whether it's physical security, you know,

571
00:36:08,680 --> 00:36:12,320
or it security. Like you know, the things

572
00:36:12,320 --> 00:36:16,080
developers have told her. Right.

573
00:36:16,080 --> 00:36:19,600
You know, and like, as a form, you know, as a former developer, like, I

574
00:36:19,600 --> 00:36:22,840
get it, like, and I know data scientists don't think about this

575
00:36:23,080 --> 00:36:26,920
generally speaking, right. Data engineers might,

576
00:36:27,480 --> 00:36:31,280
but even then, like, you know, they're, I think you said

577
00:36:31,280 --> 00:36:34,760
it earlier like it's the mindset, right? You know, you have the builder mindset,

578
00:36:35,560 --> 00:36:39,400
maybe the plumber mindset for data engineers and

579
00:36:39,400 --> 00:36:43,160
then you have kind of the attacker mindset, right. These are different ways of

580
00:36:43,160 --> 00:36:46,680
thinking. It almost cries out that you need to have

581
00:36:48,760 --> 00:36:52,040
diverse mindsets on these projects now. I mean, you always need.

582
00:36:52,600 --> 00:36:56,280
Now it's more obvious. That's why you see, yeah,

583
00:36:56,280 --> 00:36:57,240
that's why you see

584
00:36:59,880 --> 00:37:03,660
good hygiene insecurity is that you have, you do

585
00:37:03,660 --> 00:37:07,500
a threat modeling before you are going to go external. And

586
00:37:07,500 --> 00:37:11,340
that is a very different person that usually guides that. It's, it's exactly what you

587
00:37:11,340 --> 00:37:14,820
said, Frank. They have, they have the. I'm going to bring to you

588
00:37:15,220 --> 00:37:18,660
this, what you might feel is a very, very big edge case.

589
00:37:19,140 --> 00:37:22,980
But if there's a probability can happen, you have to consider

590
00:37:22,980 --> 00:37:26,790
it and you have to, you have to think about it. And that's back

591
00:37:26,790 --> 00:37:30,230
to something we talked about earlier. When you have the attackers using

592
00:37:30,230 --> 00:37:33,470
AI, they can explore the

593
00:37:34,190 --> 00:37:37,870
corners and these edge cases so much easier.

594
00:37:38,830 --> 00:37:42,189
And if they find one. And it could be very

595
00:37:42,189 --> 00:37:46,030
unsophisticated though, because it could still be a vulnerability

596
00:37:46,110 --> 00:37:49,910
that has been around for 10 years and believe it or not,

597
00:37:49,910 --> 00:37:53,760
that's still going on that the ultimate way they got in

598
00:37:54,080 --> 00:37:57,280
was a very old unpatched

599
00:37:57,440 --> 00:38:01,200
resource that just happened to get exposed. But it was, it was the

600
00:38:02,400 --> 00:38:06,240
reason other term in security, the lateral movement of the bad guy, that they're

601
00:38:06,240 --> 00:38:09,600
just, they're just moving laterally to investigate

602
00:38:09,840 --> 00:38:13,360
different parts and they found something and they got in that way

603
00:38:13,360 --> 00:38:17,200
and back, back to the thing. Just have to be right once and

604
00:38:17,360 --> 00:38:20,970
poor defenders have to be right 100% of the time and they

605
00:38:20,970 --> 00:38:24,370
won't be. So that's why taking a risk approach

606
00:38:24,530 --> 00:38:28,210
is the way, is the way to go. Because no matter

607
00:38:28,210 --> 00:38:31,330
what your size of the company, you got to consider your budget,

608
00:38:31,810 --> 00:38:35,410
how many people you have, the skill set of those people.

609
00:38:36,290 --> 00:38:39,490
And this is where I think AI can really assist the

610
00:38:39,970 --> 00:38:43,490
defenders is that it can add some of that

611
00:38:43,490 --> 00:38:46,850
expertise and some of that, you know, vigilance

612
00:38:46,930 --> 00:38:50,490
that's on 24. 7 in ways that people

613
00:38:50,490 --> 00:38:53,250
can't. But they got to bring it to people and the people can make the

614
00:38:53,250 --> 00:38:56,610
judgment call. Because if the AI had its way, I mean,

615
00:38:57,250 --> 00:39:00,850
Frankie kind of said this too, that the most secure thing is to

616
00:39:00,850 --> 00:39:04,450
shut the whole thing down and not let customers access it.

617
00:39:04,610 --> 00:39:07,890
You know, and you don't want that because that, that's your business, you know. That

618
00:39:07,890 --> 00:39:11,530
kind of defeats the purpose. Yeah, exactly, exactly.

619
00:39:11,530 --> 00:39:15,230
So a risk based approach is super important. And, and it is

620
00:39:15,230 --> 00:39:18,950
about then just, you know, you judging how much

621
00:39:18,950 --> 00:39:22,670
risk you want to take and your board wants to take and you

622
00:39:22,670 --> 00:39:25,870
know, and the CEO wants to take and the business people want to take and

623
00:39:25,870 --> 00:39:29,150
then, and then applying that and making sure that,

624
00:39:29,870 --> 00:39:32,510
you know, it matches your business.

625
00:39:34,110 --> 00:39:37,550
And so that's, you know, that, that's a lot of the game. Right.

626
00:39:38,590 --> 00:39:41,270
That makes a lot of sense. So what is your. I'm sorry, candid. Go ahead.

627
00:39:41,270 --> 00:39:44,920
So what would trigger the shift inside an organization

628
00:39:45,400 --> 00:39:49,080
from reactive security to risk aligned decision

629
00:39:49,160 --> 00:39:52,880
making? You

630
00:39:52,880 --> 00:39:56,440
know, oftentimes, unfortunately,

631
00:39:56,440 --> 00:39:59,720
it's you get hacked and

632
00:40:00,280 --> 00:40:03,960
you, a lot of times also, unfortunately, you bring

633
00:40:03,960 --> 00:40:07,800
in new leadership who understand that their charter is to

634
00:40:07,800 --> 00:40:11,520
come in and change the culture a bit, you know,

635
00:40:11,600 --> 00:40:15,280
from that. Now existing leadership can certainly do that, but

636
00:40:15,280 --> 00:40:19,000
whether they're given enough chance to, I

637
00:40:19,000 --> 00:40:22,680
don't know, I, you know, it's. All fun and games

638
00:40:22,680 --> 00:40:26,400
until somebody gets hurt. Right. Like, and you know, and I think that

639
00:40:28,240 --> 00:40:32,040
if you'd never had a problem before and then it suddenly

640
00:40:32,040 --> 00:40:35,830
happens. Right. I don't think it's, there's

641
00:40:35,830 --> 00:40:38,710
a joke, it's a bit of a gallows humor type thing where

642
00:40:40,230 --> 00:40:44,070
like clockwork, within 24 hours of a major breach of a major company. Right.

643
00:40:44,550 --> 00:40:47,790
What do you see? Job listings for

644
00:40:47,790 --> 00:40:51,590
cybersecurity? Okay. There was a major, I

645
00:40:51,590 --> 00:40:54,230
think it was one of the major hotel chains. I think you all know who

646
00:40:54,230 --> 00:40:56,350
we're talking about. I don't want to, I don't want to name Anyone by name,

647
00:40:56,350 --> 00:40:59,990
I don't want to get sued. But you know, literally

648
00:40:59,990 --> 00:41:03,590
like within a week, you know, there was like two pages of job

649
00:41:03,590 --> 00:41:06,710
listings for, you know, some flavor of

650
00:41:06,710 --> 00:41:09,270
cybersecurity or security analysis.

651
00:41:10,550 --> 00:41:13,870
And it's unfortunate that in many

652
00:41:13,870 --> 00:41:17,710
organizations the security leader is kind of

653
00:41:17,710 --> 00:41:21,030
set up to be the scapegoat when something like that happens. When

654
00:41:21,430 --> 00:41:25,030
in fact, you know, you could be doing all the right things.

655
00:41:25,510 --> 00:41:29,320
And I don't know, you guys probably know the term

656
00:41:29,320 --> 00:41:32,800
too. The, a black swan event kind of happens. Which, right, which,

657
00:41:33,040 --> 00:41:36,800
you know, we know it, everybody knows it if they travel. Because

658
00:41:37,120 --> 00:41:40,720
how often have we caught someone who has the explosive in their shoes

659
00:41:40,960 --> 00:41:44,800
getting on an airplane? It's never happened since the first time.

660
00:41:45,120 --> 00:41:48,400
That, and that was a black swan event. And yet we

661
00:41:48,720 --> 00:41:52,320
designed our whole security, a lot of our security around

662
00:41:52,400 --> 00:41:55,950
that and that. And it should be done around the major,

663
00:41:56,190 --> 00:41:58,910
the risks. And if you think about

664
00:41:59,870 --> 00:42:03,390
it in that way, really if you've

665
00:42:03,390 --> 00:42:07,230
traveled internationally, especially in places that they really have a risk,

666
00:42:07,630 --> 00:42:10,670
they will often randomly pick a plane,

667
00:42:11,230 --> 00:42:14,750
get everybody off, look at all the baggage. But it's a

668
00:42:14,750 --> 00:42:17,630
random kind of thing that happens

669
00:42:18,590 --> 00:42:22,430
rather than kind of a systemic way of going through it that becomes

670
00:42:23,110 --> 00:42:26,470
kind of wrote and it, you know, and, and people learn how to defeat it,

671
00:42:26,470 --> 00:42:29,990
you know, in some ways. And, and that happens in cyber security

672
00:42:30,070 --> 00:42:32,950
all the time. You know, you gotta, you gotta really be

673
00:42:33,830 --> 00:42:37,630
doing that. That's why doing like practice, you know,

674
00:42:37,630 --> 00:42:41,190
it's, it's a, it's a real important thing to do what we call

675
00:42:41,270 --> 00:42:44,750
tabletop exercises in this because you have to

676
00:42:44,750 --> 00:42:47,910
pretend like you just got hacked. What, what do you do

677
00:42:48,480 --> 00:42:52,200
from the lowest level analyst all the way up to

678
00:42:52,200 --> 00:42:55,840
the board. Do they, do they know what to do? Because,

679
00:42:55,920 --> 00:42:59,520
and there's a lot of regulations now within like 24 hours, they have to,

680
00:42:59,520 --> 00:43:03,160
or 72 hours of detecting it, they have to, they're on the

681
00:43:03,160 --> 00:43:06,840
hook to claim it. I forget what that law is called. Yeah,

682
00:43:06,840 --> 00:43:10,320
that's right there. If you're a public company, you have to disclose

683
00:43:10,560 --> 00:43:14,000
and typically you don't have any idea yet

684
00:43:14,680 --> 00:43:18,520
what, how that's happened and yet you have to disclose it's

685
00:43:18,520 --> 00:43:21,240
happened and it's, you know, so,

686
00:43:22,360 --> 00:43:26,040
so yeah, there's, there's a lot of risk to the organization

687
00:43:26,280 --> 00:43:30,120
that, that this, this presents. And so that's why having,

688
00:43:30,840 --> 00:43:34,680
thinking about it that way, doing exercises, you

689
00:43:34,680 --> 00:43:38,520
know, it is a new world. I'll say. I, I'm

690
00:43:38,520 --> 00:43:42,280
a big believer in what I'm going to call situational security where

691
00:43:43,050 --> 00:43:46,210
And I mentioned this before, you just got to know your situation and if the

692
00:43:46,210 --> 00:43:49,730
stakes are high and you have a big security team and you better be

693
00:43:49,730 --> 00:43:53,250
practicing these things, you better have done, you know,

694
00:43:53,250 --> 00:43:57,050
multi layers of security. But if you're a small team and you only have a

695
00:43:57,050 --> 00:44:00,850
couple people on it, you've got to kind of think of what your crown jewels

696
00:44:00,850 --> 00:44:04,570
are. Go protect those first and let the other

697
00:44:04,570 --> 00:44:07,770
stuff go because who cares who's on your guest network?

698
00:44:08,250 --> 00:44:11,980
You know, it's like you got to let that go, maybe make sure that,

699
00:44:11,980 --> 00:44:14,380
make sure your guest network is not tied to your internal network.

700
00:44:17,420 --> 00:44:20,940
And I think these days you have to really look at access

701
00:44:21,740 --> 00:44:25,500
because so much. Everyone's in the cloud with

702
00:44:25,500 --> 00:44:29,140
a lot of their infrastructure these days and you can tell a lot

703
00:44:29,140 --> 00:44:32,620
by, so don't over privilege people to have access to things

704
00:44:32,860 --> 00:44:36,100
and do that. So you have to look at those kinds of things. You do

705
00:44:36,100 --> 00:44:39,370
have to look at your, you know, it goes without saying,

706
00:44:39,690 --> 00:44:43,450
look at, look at your resources and I'm going to use that term broadly,

707
00:44:43,530 --> 00:44:47,130
your assets that you have because you have to know about them.

708
00:44:47,370 --> 00:44:50,410
So having some protection on those assets is super critical as well.

709
00:44:51,050 --> 00:44:54,850
And I'm an old appsec guy, so yes, you, and you know, Frank,

710
00:44:54,850 --> 00:44:58,570
you mentioned you got to have your applications

711
00:44:58,650 --> 00:45:02,410
that are actually performing much of the business these days. You have to,

712
00:45:02,660 --> 00:45:06,220
you have to know what your vulnerabilities are and you've got to plug the big

713
00:45:06,220 --> 00:45:09,780
holes in that. But from there really

714
00:45:09,780 --> 00:45:13,140
can't stop the bad guys. But you can at least stop,

715
00:45:13,460 --> 00:45:17,300
stop the amateur bad guys. Right. Well, and, and

716
00:45:17,300 --> 00:45:21,140
they're going to look around. So you can, if your bar is

717
00:45:21,140 --> 00:45:24,700
higher than the next guy, as we know. You know, I know

718
00:45:24,700 --> 00:45:28,420
that's all these adages of, you know, running faster than a bear or

719
00:45:28,650 --> 00:45:31,730
the next guy, you know, on the bear and all things. That'd be the second

720
00:45:31,730 --> 00:45:35,410
slowest. That's right. And it is true, you know, you can, you can

721
00:45:35,410 --> 00:45:38,890
dissuade a lot of attacks if you

722
00:45:39,210 --> 00:45:42,970
look like it's going to be difficult because the attackers,

723
00:45:43,450 --> 00:45:47,170
they run playbooks too, because it's easier for them, it's cheaper for

724
00:45:47,170 --> 00:45:50,930
them. It. And they'll just run playbooks. And if, if you

725
00:45:50,930 --> 00:45:54,370
thwart the playbook, they'll find someone who, who

726
00:45:54,370 --> 00:45:58,080
doesn't. And if you're not state actor, it's a, it's a criminal

727
00:45:58,080 --> 00:46:01,840
enterprise. Right. And criminals are there to make money. Right. State

728
00:46:01,840 --> 00:46:05,440
actors have different motives and different budgets.

729
00:46:05,600 --> 00:46:08,600
Yeah. They may go a lot more targeted and they're just going to wait and

730
00:46:08,600 --> 00:46:10,960
be patient. You're exactly right. But

731
00:46:12,080 --> 00:46:14,960
actually targets like. I'm sorry,

732
00:46:16,400 --> 00:46:20,240
no, I don't mean to interrupt. I was just going to say a funny story

733
00:46:20,400 --> 00:46:22,960
is when we were

734
00:46:24,480 --> 00:46:28,120
pitching our application security company

735
00:46:28,120 --> 00:46:31,960
back in 2003, we used to talk about

736
00:46:31,960 --> 00:46:32,320
how

737
00:46:34,800 --> 00:46:38,600
underfunded but patient and have all the times in the

738
00:46:38,600 --> 00:46:42,000
world the hackers are, we were kind of

739
00:46:42,000 --> 00:46:44,880
dismissive of no state would ever

740
00:46:45,440 --> 00:46:49,080
hack another state's assets because it start a

741
00:46:49,080 --> 00:46:52,640
war. And at the time that was really, that was the thinking.

742
00:46:52,880 --> 00:46:56,160
I mean, how quaint does that sound today when we all know it's like, oh,

743
00:46:57,040 --> 00:47:00,880
that's a Russian hacker group. You know, it's like we just kind of go,

744
00:47:00,880 --> 00:47:04,080
oh, of course it was. It's like, oh my gosh. Yeah.

745
00:47:04,480 --> 00:47:07,880
Well, it's also, I think in terms of geopolitics become a real

746
00:47:07,880 --> 00:47:11,560
equalizer. Right. Because a nation state like North Korea can go toe to

747
00:47:11,560 --> 00:47:15,200
toe with the United States. Right. Whereas in a

748
00:47:15,200 --> 00:47:18,880
conventional war really wouldn't work out well for them. You know what I mean?

749
00:47:18,880 --> 00:47:22,720
It's an interesting, yeah, it is interesting.

750
00:47:23,040 --> 00:47:26,760
We have really good hackers ourselves in the

751
00:47:26,760 --> 00:47:30,560
United States, right? Oh, I'm sure we do. It's, it's, it, you know,

752
00:47:31,040 --> 00:47:34,720
I mean I, you know, you kind of hope that but,

753
00:47:34,720 --> 00:47:38,320
but you are right, like a North Korean thing like, like we've seen

754
00:47:39,760 --> 00:47:43,520
they can use these deep fakes to infiltrate in ways

755
00:47:43,520 --> 00:47:47,160
that, you know, because of the work at home thing how, how they

756
00:47:47,160 --> 00:47:50,830
can get employees hired in some of these places

757
00:47:51,070 --> 00:47:54,670
with the expressed intent of, you know, stealing

758
00:47:54,670 --> 00:47:57,870
things, you know, from, from those organizations and it's,

759
00:47:58,910 --> 00:48:02,510
yeah, it's, it's a new world. It's pretty wild. Yeah. I mean when you think

760
00:48:02,510 --> 00:48:05,910
about it like in, and you know, and it's not, not saying that the United

761
00:48:05,910 --> 00:48:09,350
States doesn't have good hackers. I'm sure, I'm sure we have among the best. I

762
00:48:09,350 --> 00:48:13,070
mean, maybe the best. But it's like a

763
00:48:13,070 --> 00:48:16,390
baseball team, right? Like, you know, obviously there are some baseball teams that are going

764
00:48:16,390 --> 00:48:19,870
to be better than others, right. And it's going to be kind of like the

765
00:48:19,870 --> 00:48:23,670
smaller town that doesn't have the budget to pay for this. Rock stars. Same

766
00:48:23,670 --> 00:48:27,030
with football, right? Whatever sport your thing is, right. You know, for me, I'm a

767
00:48:27,030 --> 00:48:30,350
Yankees fan, although the Yankees have not had a good run of late. But

768
00:48:30,350 --> 00:48:34,190
historically they have been kind of the top. But you know, you

769
00:48:34,190 --> 00:48:37,990
can definitely tell like nation states can be all in like the same league

770
00:48:38,070 --> 00:48:41,720
because they do have more or Less the same capacity in terms of. They're

771
00:48:41,720 --> 00:48:45,080
not, they're not in it for the money per se. Like, you know what I

772
00:48:45,080 --> 00:48:48,800
mean? They're not, you know, they, they. Because they're a

773
00:48:48,800 --> 00:48:52,120
nation state, you know, they can harbor. They can harbor themselves and not

774
00:48:52,120 --> 00:48:55,720
prosecute. You know, they have certain more advantages than your average criminal gang.

775
00:48:56,680 --> 00:49:00,480
Oh yeah. I mean. And well funded. Right. I mean, that's. Money's not

776
00:49:00,480 --> 00:49:04,160
an issue. Yeah, right. It's it. It

777
00:49:04,160 --> 00:49:07,800
puts to be a formative adversary.

778
00:49:08,420 --> 00:49:12,020
And that's why places

779
00:49:12,020 --> 00:49:15,540
like Mandy and that come out with threat reports.

780
00:49:16,420 --> 00:49:20,180
They talk about these actor groups, but you could see moves of

781
00:49:20,180 --> 00:49:22,740
actor groups as well,

782
00:49:23,780 --> 00:49:27,540
changing their tactics and techniques. Again, one of the

783
00:49:27,540 --> 00:49:31,340
more interesting things that happened. I think we could talk about it

784
00:49:31,340 --> 00:49:33,620
because it is public. But if you remember

785
00:49:35,770 --> 00:49:38,250
maybe a year and a half ago, maybe it was two years ago that, that

786
00:49:38,810 --> 00:49:42,570
mgm, you know, casinos guy. Oh yeah, the resort. Remember that?

787
00:49:42,810 --> 00:49:46,450
And it shut. It shut down two casinos with

788
00:49:46,450 --> 00:49:49,610
ransomware. The actors that did that.

789
00:49:50,810 --> 00:49:54,530
It goes back to what's come full circle where it used to be the adversaries,

790
00:49:54,530 --> 00:49:58,090
where it used to be called script kiddies were basically kids

791
00:49:58,410 --> 00:50:02,190
that want to just cause disruption. Well, this is, this is actually a

792
00:50:02,190 --> 00:50:05,870
more sophisticated. It ends up. This group was just a more sophisticated thing of that.

793
00:50:05,870 --> 00:50:09,670
They. Yeah, it was ransomware, but they weren't actually out there

794
00:50:09,670 --> 00:50:11,950
just for the money. They just wanted to do it. They just wanted to see

795
00:50:11,950 --> 00:50:15,390
if they could shut down a casin. And it's crazy that that

796
00:50:15,550 --> 00:50:19,230
that's like that and they still got away with a bunch of crypto

797
00:50:20,190 --> 00:50:22,270
money. But. But it,

798
00:50:23,870 --> 00:50:27,490
you know, it just shows that even like

799
00:50:29,330 --> 00:50:33,090
those, even those hackers could then stand on the shoulders of

800
00:50:33,090 --> 00:50:36,810
all this technology that's being hopefully built for

801
00:50:36,810 --> 00:50:40,650
good and stuff too. But they can use that. And now

802
00:50:40,650 --> 00:50:43,810
you can generate. I know the LLMs.

803
00:50:44,770 --> 00:50:48,090
Back to the subject. You could ask it to

804
00:50:48,090 --> 00:50:51,690
generate malware for you and it'll at

805
00:50:51,690 --> 00:50:54,610
first say no, but if you could trick it, it'll say yes and it'll do

806
00:50:54,610 --> 00:50:58,130
it. And then you can. Didn't that happen recently where there were. State actors

807
00:50:58,850 --> 00:51:02,370
with cl. Yeah, anthropic.

808
00:51:03,330 --> 00:51:06,530
That they were. That. That that plot had been used

809
00:51:07,010 --> 00:51:10,730
and they're, you know, I, I think anthropic really looks at the

810
00:51:10,730 --> 00:51:14,450
safety of what they're doing and stuff too. So they, that's why they disclosed it.

811
00:51:14,450 --> 00:51:18,130
But. And they filled that hole. But it was. It wasn't that hard.

812
00:51:18,130 --> 00:51:21,610
You know, all they did was say, oh no, I'm a Researcher. And

813
00:51:21,690 --> 00:51:25,530
I'm doing an ethical. Yep. It

814
00:51:25,530 --> 00:51:29,130
was not an ethical. How would you. I mean, you've been in the

815
00:51:29,130 --> 00:51:32,530
AppSec before. It was called cyber security. Was called

816
00:51:32,530 --> 00:51:36,010
AppSec or application security. But, you know,

817
00:51:36,330 --> 00:51:39,810
if somebody told you back when you said, you know, no nation state would do

818
00:51:39,810 --> 00:51:43,370
this, right. That, you know, all you had to do is trick a computer into

819
00:51:43,370 --> 00:51:46,730
giving you, like, talk to a computer and tell it you're a research. Like, how

820
00:51:46,730 --> 00:51:50,290
unreal is that? Like, I don't know. I'm doing this for research. Like, oh,

821
00:51:50,290 --> 00:51:54,050
okay. Like, yeah, you know. Yeah, pretty, pretty,

822
00:51:54,050 --> 00:51:56,810
pretty unreal because it was so manual before,

823
00:51:57,930 --> 00:52:01,730
you know where. But again, you know, it gets back

824
00:52:01,730 --> 00:52:05,450
to the thing we talked about earlier. It's like there are builders of the

825
00:52:05,450 --> 00:52:08,970
world that can't imagine someone wanting to destroy,

826
00:52:09,370 --> 00:52:13,090
you know, this beautiful building that's, that's been, that's been

827
00:52:13,090 --> 00:52:16,800
built. And then there's people that. All they think about is, how can I

828
00:52:16,800 --> 00:52:20,560
find a weakness in that building and take it, either take it down or just

829
00:52:20,560 --> 00:52:23,800
gain access and that. Right, that's, that's, that's what's around.

830
00:52:23,960 --> 00:52:27,640
Yeah. I mean, that keeps on. If you're in cyber security, that, that,

831
00:52:27,640 --> 00:52:31,320
that, that keeps the lights on for sure. Because there's always,

832
00:52:31,639 --> 00:52:35,400
always, there's always work to do to help the defenders.

833
00:52:36,280 --> 00:52:39,280
Yeah. So you have something to defend. It was like going back to medieval times,

834
00:52:39,280 --> 00:52:42,960
right. Like, you had the kings, but you had a pretty large class of

835
00:52:42,960 --> 00:52:46,360
knights, you know, that would have to do defending and. Or

836
00:52:46,840 --> 00:52:49,360
I forget what the people were called, but they would stand on the walls and,

837
00:52:49,360 --> 00:52:52,200
like shoot arrows and catapults and stuff like. Yeah.

838
00:52:54,280 --> 00:52:58,120
And you design the moat is because of that. And then

839
00:52:58,120 --> 00:53:01,600
the ways you get into the city, you know, has got

840
00:53:01,600 --> 00:53:05,120
traps in it, you know, and we would

841
00:53:05,120 --> 00:53:08,680
liken that to a honey pot, you know, I mean, there's lots of, lots of.

842
00:53:09,410 --> 00:53:13,090
And then Trojan horse was originally a Trojan horse.

843
00:53:13,650 --> 00:53:17,290
That's right. And it's a battle. Right. And so I think right

844
00:53:17,290 --> 00:53:21,010
now with AI, the attackers have a bit of the upper hand

845
00:53:21,090 --> 00:53:24,210
because we just don't know how they're using it.

846
00:53:24,930 --> 00:53:28,770
But, you know, there'll be tools and there already

847
00:53:28,850 --> 00:53:32,210
are. I mean, if you're a cybersecurity company and you don't have

848
00:53:32,690 --> 00:53:36,480
some AI assistance to help,

849
00:53:36,480 --> 00:53:39,600
either with the scope or the breadth or the speed,

850
00:53:40,080 --> 00:53:43,680
you know, it's. You see that and that. But that's on the detection

851
00:53:43,680 --> 00:53:46,480
end, and sometimes that's too late. I mean,

852
00:53:47,280 --> 00:53:51,120
I hope that the industry moves to some prevention. And then it is

853
00:53:51,120 --> 00:53:54,920
about the building of the moats or the maze that they

854
00:53:54,920 --> 00:53:58,240
have to go through or something like that. And I think that's an important

855
00:53:59,200 --> 00:54:02,720
balance that has to be maintained by enterprises today

856
00:54:03,770 --> 00:54:07,610
to. To make sure that they mitigate the risk. No, that's a

857
00:54:07,610 --> 00:54:08,330
good way to put it.

858
00:54:12,410 --> 00:54:14,570
Any question? We're getting close to the top of the hour, so I want to

859
00:54:14,570 --> 00:54:18,050
be respectful of your time. Any questions? Candace? Sorry, I. No,

860
00:54:18,050 --> 00:54:21,690
honestly, like, this has been a fantastic, fantastic interview.

861
00:54:21,690 --> 00:54:25,210
It's been incredibly enlightening. Like, so much to think about,

862
00:54:25,450 --> 00:54:29,170
you. Know, makes me want to change all my passwords. Right,

863
00:54:29,170 --> 00:54:32,650
right. Well, you know, password 1, 2, 3, nobody. No, you can't. That's not secure

864
00:54:32,650 --> 00:54:36,390
anymore, you know. Well, even, you know, it's interesting

865
00:54:36,470 --> 00:54:39,670
because talk about checkbox compliance versus real

866
00:54:40,390 --> 00:54:43,110
security. Even as we were setting up the company,

867
00:54:44,070 --> 00:54:47,710
little old us, you know, we go to, and we're

868
00:54:47,710 --> 00:54:50,550
needing help because we're wanting to become compliant to some of these

869
00:54:51,030 --> 00:54:54,830
bars that are out there, like SoC2, if you've heard

870
00:54:54,830 --> 00:54:57,430
of that. It's a compliance standard

871
00:54:58,390 --> 00:55:02,200
for trustworthiness of. Of companies like us who might have your day.

872
00:55:02,200 --> 00:55:06,040
There's a massive Alphabet soup there. There is,

873
00:55:06,040 --> 00:55:09,800
there is. And, but like, password policy was really interesting

874
00:55:09,800 --> 00:55:12,840
because what. We were using some AI to help us

875
00:55:13,480 --> 00:55:17,240
in that, and it came back with password policy of. Oh,

876
00:55:17,240 --> 00:55:21,080
yeah, you know, like, change your password every two weeks. Well, that's

877
00:55:21,080 --> 00:55:24,080
been. That might have been state of the art a couple of years ago, but

878
00:55:24,080 --> 00:55:27,460
that's not what you do today. Today, you know, it's about

879
00:55:27,780 --> 00:55:31,140
length and scrambling, and we have these things called password

880
00:55:31,140 --> 00:55:34,940
managers that allow us to do that rather than

881
00:55:34,940 --> 00:55:38,500
us remembering something and. Yeah, even that.

882
00:55:39,540 --> 00:55:41,620
Dynamics. Yes, those can get hacked.

883
00:55:43,380 --> 00:55:44,900
A recent breach on one of those.

884
00:55:46,980 --> 00:55:50,780
You know, there's. I. You know, I don't know, one that was

885
00:55:50,780 --> 00:55:54,100
really bad publicly. There has been in the past, for sure. All right.

886
00:55:54,390 --> 00:55:58,150
Um, yeah, so. But you're still, I think, Correct me if I'm wrong, but I

887
00:55:58,150 --> 00:56:01,870
still think you're safer with a password Manager without

888
00:56:01,870 --> 00:56:05,510
it 100% you. It's.

889
00:56:05,510 --> 00:56:08,710
It's because you want. You want long length,

890
00:56:09,430 --> 00:56:12,950
jumbled, you know, kinds of things that just aren't easy,

891
00:56:13,270 --> 00:56:15,190
easy for the attacker. So.

892
00:56:16,790 --> 00:56:20,630
Yeah, and then you change your master password of that manager

893
00:56:21,090 --> 00:56:23,930
frequently. That's one where you, you. And again, you still want it to be long

894
00:56:23,930 --> 00:56:27,770
and. Right, right, right. Long and complicated. Remember, one long

895
00:56:27,770 --> 00:56:31,610
and complicated thing. Well, I'm, I'm glad

896
00:56:31,610 --> 00:56:35,290
that we got to kind of talk yeah, awesome. It's been

897
00:56:35,290 --> 00:56:38,170
great. Yeah, it's been great. Where can folks find out more about you and your

898
00:56:38,170 --> 00:56:41,930
company? So I think today, you know, like I said, we're

899
00:56:41,930 --> 00:56:45,250
in stealth, but eventually, please follow

900
00:56:45,330 --> 00:56:49,050
Pulse Security AI will be coming out of stealth, you know, in

901
00:56:49,050 --> 00:56:52,790
the, in the probably new year to mid

902
00:56:52,790 --> 00:56:56,390
year kind of thing. But also we started a community

903
00:56:56,710 --> 00:57:00,390
of security professionals and we as

904
00:57:00,390 --> 00:57:02,670
just a networking organization, we call that

905
00:57:02,670 --> 00:57:05,910
securityimpactcircle.org and there

906
00:57:07,110 --> 00:57:10,750
we have blogs. We want to have

907
00:57:10,750 --> 00:57:14,590
people talking about this prevention versus detection or even for

908
00:57:14,590 --> 00:57:18,400
the security leader, about risk and how they should manage

909
00:57:18,400 --> 00:57:22,240
things and best practices that they have together. So

910
00:57:22,240 --> 00:57:26,080
yeah, we have a site, securityimpactcircle.org that is a great place for people

911
00:57:26,080 --> 00:57:29,760
to go and eventually, you know, you'll get to us through that as

912
00:57:29,760 --> 00:57:33,599
well. Cool. Awesome. Well, I'll let

913
00:57:33,599 --> 00:57:37,280
our AI finish the show. And that's a wrap on this

914
00:57:37,280 --> 00:57:40,960
episode of Data Driven. Big thanks to Mike Armistead for

915
00:57:40,960 --> 00:57:44,740
reminding us that while AI may be the future, security breaches

916
00:57:44,740 --> 00:57:48,500
are very much the present. Remember, the attackers only have

917
00:57:48,500 --> 00:57:51,940
to be right once. So maybe don't make your password password

918
00:57:52,260 --> 00:57:56,060
until next time. Stay curious, stay secure, and for

919
00:57:56,060 --> 00:57:59,420
the love of data, please update your firmware. Cheers for

920
00:57:59,420 --> 00:58:01,700
listening. Now go change that password.