WEBVTT

1
00:00:00.542 --> 00:00:01.960
Generative AI, or GenAI,

2
00:00:01.960 --> 00:00:04.170
is advancing businesses worldwide,

3
00:00:04.170 --> 00:00:06.047
but as companies push to adopt it,

4
00:00:06.047 --> 00:00:08.883
they also increase exposure
to technology risks

5
00:00:08.883 --> 00:00:11.052
and broaden potential attack surfaces.

6
00:00:11.052 --> 00:00:14.180
When it comes to GenAI risk,
how can companies be prepared?

7
00:00:15.056 --> 00:00:17.392
Implementing GenAI means determining

8
00:00:17.392 --> 00:00:19.686
the best use case for scaling up,

9
00:00:19.686 --> 00:00:21.521
but companies also need to ensure

10
00:00:21.521 --> 00:00:24.899
that the GenAI models that
they build, buy, and implement

11
00:00:24.899 --> 00:00:27.610
have appropriate guardrails
that ensure data protection,

12
00:00:27.610 --> 00:00:29.571
privacy, and responsible use.

13
00:00:30.572 --> 00:00:32.407
First, the good news.

14
00:00:32.407 --> 00:00:35.452
You don't have to do security
differently for GenAI,

15
00:00:35.452 --> 00:00:38.246
but there are three things
that you need to be aware of

16
00:00:38.246 --> 00:00:39.873
as you look at these solutions.

17
00:00:40.749 --> 00:00:44.294
One, look over the landscape
of your GenAI application

18
00:00:44.294 --> 00:00:46.254
and be aware of how attackers

19
00:00:46.254 --> 00:00:47.922
could use it to their advantage.

20
00:00:48.882 --> 00:00:52.469
Two, ask yourself, how can you use GenAI

21
00:00:52.469 --> 00:00:54.429
to make your security teams better,

22
00:00:54.429 --> 00:00:56.890
both from a training and
a tooling perspective?

23
00:00:57.849 --> 00:01:00.769
And three, integrate GenAI considerations

24
00:01:00.769 --> 00:01:03.563
into your overall
enterprise security program.

25
00:01:03.563 --> 00:01:05.565
This includes third parties.

26
00:01:05.565 --> 00:01:07.567
Models in development need to be made safe

27
00:01:07.567 --> 00:01:09.486
against AI-specific threats,

28
00:01:09.486 --> 00:01:12.530
such as vulnerabilities in
the algorithm supply chain,

29
00:01:12.530 --> 00:01:15.158
and the integrity of the training data.

30
00:01:15.158 --> 00:01:17.118
The security of vendor and partner systems

31
00:01:17.118 --> 00:01:20.580
must be fully understood to
evaluate potential risks.

32
00:01:20.580 --> 00:01:24.167
From there, implementing these
practices are the best way

33
00:01:24.167 --> 00:01:26.461
to make sure you're
protecting your own data

34
00:01:26.461 --> 00:01:28.421
and your customers' as well.

35
00:01:28.421 --> 00:01:31.299
If you have a security
mindset from the beginning,

36
00:01:31.299 --> 00:01:33.218
you'll enjoy the benefits of GenAI

37
00:01:33.218 --> 00:01:34.844
while lowering your company's risk,

38
00:01:34.844 --> 00:01:37.180
and minimizing the
potential for tech debt.