Up next in 10
Shlomo Kramer, founder of Cato Networks has told City AM London is a well-respected location for AI talent, and plans to open a hub in the capital.
The AI tycoon plans to expand headcount and bring all existing London employees under one roof.
Simon Hunt has the story.
Get more of City AM 👇
🌐 http://www.cityam.com
X(formerly Twitter): http://twitter.com/CityAM
Facebook: http://www.facebook.com/cityam
Instagram: http://www.instagram.com/city_am
LinkedIn: http://www.linkedin.com/company/cityam
Show More Show Less View Video Transcript
0:00
So, I'm here in the studio with
0:01
Kramer, who's the co-founder of Cato
0:04
Networks.
0:05
Um first of all, you've got an
0:06
announcement uh today about uh the
0:09
company's expansion uh into London. So,
0:11
I wonder if you could tell us a bit
0:12
about that and the and the motivations
0:14
behind that decision.
0:15
>> Yeah, we we've announced that we've
0:17
opened a new development center center
0:19
around AI in London. This is our third
0:23
uh development center. Obviously, we
0:24
started in Tel Aviv, where our
0:27
headquarters is. Uh we've expanded to
0:29
Prague. Um and now uh
0:33
London is going to be really focused
0:34
around our push
0:37
uh for AI security, which is obviously
0:41
uh going to be uh
0:42
a central uh effort and focus area for
0:46
us in the next foreseeable future.
0:48
>> It does seem like um at least f- for
0:51
London and especially for King's Cross,
0:54
uh it has become an area of sort
0:56
international renown for for AI. Uh you
0:59
know, we've got uh OpenAI opening an
1:01
office there. Uh Anthropic, DeepMind is
1:04
based there, of course. So, is that Was
1:06
that part of the motivation that you
1:07
want to be part of that ecosystem?
1:09
>> Uh absolutely. So, so where London has
1:13
is building a reputation for a good pool
1:16
of AI expertise.
1:18
And that was part of the uh the metrics
1:21
of considerations of where do we uh
1:24
start our next uh development center.
1:27
>> And we've also heard, I mean,
1:28
anecdotally that there are challenges
1:31
for firms moving into London securing a
1:34
decent high-quality office space. Is
1:36
that something you've encountered or
1:38
>> I It didn't propagate to me.
1:40
>> Okay.
1:41
>> So, I have to check.
1:43
>> Um in terms of sort of the the the wider
1:46
debates about um AI and uh
1:49
uh cybersecurity,
1:51
uh
1:51
where do you um stand on that? Because I
1:53
I you know, I've definitely heard
1:56
uh that a sort of bullish and bearish
1:58
case made, you know, one being that
2:01
obviously this this threat powered by by
2:03
AI is bigger than ever, but at the same
2:05
time that the the counter
2:08
or the defense capability is now bigger
2:10
than ever. So, which comes out on top in
2:12
that in that battle?
2:13
>> First of all, I have to say that this is
2:15
nothing new.
2:18
Cyber security and I've been in cyber
2:20
security since the early '90s.
2:22
Uh founder of Check Point, so I've been
2:26
since the beginning, basically. It has
2:28
always been a Red Queen race in the
2:30
sense that you need to run as fast as
2:33
you can in order to stay the same place.
2:34
So, the attackers get the advantage and
2:38
then defenders
2:40
get the advantage and and so on and so
2:42
forth. With the new frontier models,
2:45
there's definitely a huge step forward
2:49
uh for the attackers.
2:51
And I'll talk about that in a second.
2:54
What's
2:55
What's the advantage?
2:58
But I would say that the security
2:59
industry is showing a lot of resilience
3:02
and a lot of agility in closing that gap
3:05
quickly and responding in time
3:10
to address these challenges. And what
3:12
are the challenges?
3:14
I would say that nothing in security is
3:15
going to change.
3:17
The vulnerabilities are going to be the
3:18
same vulnerabilities. The attack vectors
3:21
are going to be the same attack vectors.
3:23
Uh the propagation of an attack is going
3:26
to stay the same. It's going to look the
3:28
same.
3:29
What's going to
3:31
change
3:32
is the security operations,
3:36
which need to be completely changed
3:39
because the attacks now are going to be
3:42
the same but in a hundred to a thousand
3:45
times faster,
3:47
uh
3:50
bigger, and more resilient
3:53
and with the current operational uh
3:56
model
3:58
uh really organizations are going to get
4:00
stampeded. Uh they are going to be
4:02
breached.
4:04
They have to
4:06
uh um address that with a new
4:08
operational model.
4:10
At Cato, we've talked a long time about
4:12
the operational uh side of security
4:15
as a key element in enabling both
4:19
business agility and uh better security.
4:23
It was always a a question of
4:25
you know, uh soft question of how much
4:28
do you need to pay, how many people you
4:30
need to hire how much work can you do,
4:33
how uh much agility can you provide the
4:37
business. In the new security operation
4:40
uh landscape, this becomes a do-or-die
4:44
security uh issue. The kind of the right
4:48
architecture
4:49
uh uh uh uh the right approach uh
4:52
across the board, whether it's on the
4:54
SOC side, whether it's on the cloud
4:57
side, uh or what Cato does on the
4:59
networking side. I can give you a quick
5:02
example which clarifies this in a
5:05
uh
5:06
in a very simple way. Say you are
5:09
a large retailer. You've got 500 5,000
5:13
uh grocery shops
5:15
uh and other locations around Europe.
5:19
And in each one you've got an appliance,
5:22
and your vendor provides now patches for
5:24
these vulnerabilities. Even
5:27
you know, a few days quickly, a few days
5:30
after they are discovered.
5:33
You are still as the CISO of that uh
5:36
um
5:37
um uh vendor
5:40
uh of that customer, need to deploy them
5:43
in 5,000 locations. That will take
5:45
weeks, if not months, and huge amounts
5:48
of effort.
5:50
Uh in a new world where critical
5:53
vulnerability will appear every day,
5:57
uh perhaps multiple times a day,
5:59
and you have to close it, otherwise
6:03
you'll get compromised, that operational
6:05
mode doesn't work.
6:07
Cato just announced this week that we
6:09
have now a genetic capabilities with the
6:12
right infrastructure, because we are
6:14
cloud service, it's not appliance,
6:16
uh
6:17
to move from CV from vulnerability
6:20
discovery
6:22
to full protection of our entire 5,000
6:25
enterprise customer installed base in 45
6:28
minutes.
6:30
This is a much better proposition than
6:34
a few weeks. So, and the same goes if to
6:36
many other areas in security. You really
6:39
need to reinvent that operational model
6:43
in order to stay secure, and that's what
6:45
this the security industry is doing
6:47
these days ferociously, I would say.
6:51
>> And and in terms of the I mean,
6:52
obviously you talked about the the
6:53
increased sort of speed and frequency of
6:55
the of the potential threat. Uh is it is
6:58
that does that partly derive from the um
7:01
an expansion in the the range of
7:03
potential um cyber attackers in the
7:06
sense that on the one hand you've got
7:08
sophisticated ones that are now becoming
7:09
more sophisticated through AI, but also
7:11
sort of unsophisticated ones suddenly
7:13
becoming sophisticated, because they now
7:15
have much more resources available to
7:17
them very cheaply.
7:18
>> It's It's even more than that. You've
7:19
got fully agentic
7:21
vulnerability researchers, which is the
7:23
first persona in every organization that
7:26
does hacking. You've got the
7:28
vulnerability researchers that find the
7:30
tools,
7:31
and then you've got the attackers that
7:35
use the toolbox in order to
7:38
do the attack. Uh uh, you know, try
7:40
this, try that, go right, go left, try
7:43
lateral movements until you will get the
7:45
impact that you're looking for.
7:47
Uh, these two persona have been
7:49
identified now
7:51
with a very sophisticated models. And
7:53
it's not only about
7:55
uh, Anthropic models or uh,
7:59
OpenAI Daybreak. The open-source Chinese
8:03
models based on, you know,
8:05
uh,
8:06
very connected uh,
8:09
sources in Anthropic
8:11
are just 4 months behind them.
8:13
So, this is coming from everywhere. Not
8:16
not it's not dependent on a single
8:18
company or a single model.
8:20
And there's no way around it. You will
8:22
have to confront that tsunami.
8:25
Um,
8:26
and that tsunami is basically you don't
8:27
need people
8:28
or you don't need only people. Now you
8:30
can replicate them a hundred times with
8:33
agents as as long as you have the
8:35
computational resources.
8:37
>> Do you think um,
8:39
uh, you know, different markets,
8:41
different industries are aware of the
8:43
scale of that threat that you've
8:44
articulated or I mean, do you think
8:46
there are particular markets or
8:48
particular industries that are a bit
8:50
ahead of the curve, others that are
8:51
behind the curve? So, obviously you're
8:53
an international business, so you must
8:55
deal with different markets in different
8:56
ways.
8:56
>> say I would say uh,
8:58
uh,
8:59
there are five stages of grief.
9:01
>> [laughter]
9:02
>> I would say that different organization
9:05
are in different stages of that uh,
9:09
recognition and process.
9:12
And you know, some are
9:14
still in denial, still some are
9:16
negotiating,
9:17
and some have accepted it and now
9:20
uh, aggressively moving forward. But as
9:23
an industry, this is a process that's
9:25
going to be
9:26
uh, accelerating over the next 12 months
9:29
by necessity.
9:30
>> Mhm. Mhm. Mhm. Okay.
9:32
Um, I guess more more widely, I mean,
9:34
you're obviously a serial investor into
9:37
um,
9:38
big you know big tech and AI uh
9:40
companies. Uh we've got three massive uh
9:44
AI IPOs coming down the the the pipeline
9:47
now in the form of OpenAI, Anthropic,
9:49
and uh SpaceX.
9:51
Uh those components those IPOs will have
9:54
a um a retail float component to it
9:57
where retail investors can participate
9:59
if they if they want to. If you were a
10:02
prospective retail investor, would you
10:04
would you invest now [laughter] or
10:06
do you think it's too late?
10:06
>> not
10:08
Ask me questions about cyber, not about
10:10
the stock market. I
10:11
>> Yeah.
10:12
>> I'm not uh an expert.
10:14
>> know to to um
10:16
that obviously there's there's there's
10:17
lots of focus on what these big
10:19
companies are doing.
10:21
Um but how how much is that does that
10:23
feed through into what you know what
10:25
else is happening in the industry with
10:26
small and medium-sized AI companies?
10:28
>> I I I would say that the investment
10:31
community not necessarily the public
10:33
investment only but also the private
10:36
investors have been going through um
10:40
um a
10:42
a deep consideration process
10:45
uh mainly governed by fear and greed
10:49
um around
10:51
um okay, so a meteor just hit Earth.
10:56
Who are the dinosaurs and who are the
10:58
mammals? Who is going to uh be extinct
11:03
and who is going to uh flourish?
11:06
Um and I I think while they were doing
11:11
it in the beginning, first of all, they
11:12
aired toward fear which is apparently a
11:16
much stronger uh
11:18
um
11:19
uh driver.
11:21
Uh and the second, they forced.
11:24
All right, and and they to large extent
11:27
we are still there. So, when you know,
11:30
uh these models just came out, it was
11:33
hey, mitosis is going to eat the
11:34
cybersecurity industry.
11:37
No, it's not.
11:39
Actually, it's cybersecurity is an
11:43
enabler for the AI journey of every
11:46
enterprise. No enterprise is going to do
11:49
to go AI do AI without having some
11:53
guardrails, without having governance,
11:55
without putting controls over this
11:59
very sensitive new capabilities.
12:03
And it's also the beneficiary
12:05
as I explained, you know, AI
12:08
unfortunately is
12:10
accelerating the threat landscape and
12:13
you need more security in order to
12:15
address that.
12:16
So logically, you can explain why cyber
12:19
is
12:20
the mammals,
12:22
but
12:25
I would say
12:27
emotionally or whatever it was very
12:30
difficult. Like at least half a year
12:32
ago. I think more and more rationality
12:34
is coming into the mix
12:37
both in terms of investment in AI as
12:40
well as in terms of investment outside
12:43
of AI in other stuff and I believe this
12:47
is going to be a normalizing process
12:49
that's going to continue.
12:50
>> Mhm. Mhm. So I guess, you know, how how
12:51
much of the So the discussion among
12:53
investors is led by whatever is
12:56
happening by a small number of massive
12:58
AI companies
12:59
and to what extent is that valuable or
13:02
important or to what extent is it sort
13:04
of a bit of a distraction for what's
13:05
going on in the rest of the industry?
13:06
>> My guess is that a little bit of both.
13:09
A little bit, you know, I've got in
13:11
topic, it's better for me to put a
13:14
billion dollar there
13:15
than in 10 other startups because
13:21
so I'm going to focus there.
13:24
Anthropic is the code name. Um
13:28
And um
13:30
yeah, so so so that's that there's a lot
13:33
of that and there's a lot of
13:35
kind of um
13:37
um disruption happening in the market in
13:39
other areas
13:41
where you ask yourself, should I invest
13:43
in the company that was started 10 years
13:45
ago before AI or is this going area
13:48
going to be disrupted by AI by a kind of
13:51
an AI native company that started 2 or 3
13:54
years ago?
13:55
And um
13:57
my guess is that the answer to the
13:59
second question is going to be a much
14:01
more limited yes
14:02
than people thought a year ago. Like
14:05
SaaS is not
14:06
dead. Let's let's
14:08
summarize that piece. All right, it's
14:11
absolutely alive.
14:12
>> Mhm. Mhm. But I guess you know, if if
14:13
you're thinking about startups
14:16
though you have to have one eye on
14:17
whether they are developing something
14:20
that will be subsumed by one of the big
14:22
AI companies in a later model release or
14:25
something like that. So that's a factor
14:28
to consider.
14:29
>> Mhm. Yeah, I my my belief that this like
14:31
any other technology in the past is
14:36
going to have a layer beneath it uh that
14:38
enables it and obviously, you know,
14:41
megawatts and data centers
14:44
um
14:44
and security
14:46
are examples.
14:47
And there's going to be layer above it
14:50
that
14:52
winners
14:53
and losers and neutrals from that new um
14:59
layer and the the layer of the models is
15:03
not going to do
15:04
everything to every business process and
15:07
every use case in every industry.
15:11
Um just not the case. It's a new
15:14
computational model
15:16
but it's not going to eat software.
15:19
>> Mhm. Mhm. Mhm. Okay.
15:21
All right. Well, we'll leave it there
15:22
then. Thanks thanks very much for for
15:24
coming in. It was a pleasure to speak to
15:25
you.
15:25
>> Same here.
News