WEBVTT

1
00:00:00.510 --> 00:00:02.580
<v ->So everyone across your company</v>

2
00:00:02.580 --> 00:00:07.140
is probably excited to experiment with generative AI.

3
00:00:07.140 --> 00:00:10.560
And we are seeing CEOs acting fast

4
00:00:10.560 --> 00:00:14.460
to protect their companies from the business risks

5
00:00:14.460 --> 00:00:16.470
of these experiments.

6
00:00:16.470 --> 00:00:21.240
As everyone knows by now, generative AI,

7
00:00:21.240 --> 00:00:23.190
when used without effective guidelines

8
00:00:23.190 --> 00:00:25.230
and mitigation strategies,

9
00:00:25.230 --> 00:00:29.520
puts your company at risk of copyright infringement,

10
00:00:29.520 --> 00:00:31.350
leaks of proprietary data,

11
00:00:31.350 --> 00:00:35.133
lower credibility from low quality content and more.

12
00:00:36.210 --> 00:00:40.320
Many of these risks are relatively easy to mitigate

13
00:00:40.320 --> 00:00:44.160
by setting clear guidelines around use of generative AI

14
00:00:44.160 --> 00:00:46.260
and enacting responsible norms,

15
00:00:46.260 --> 00:00:51.000
such as setting up review boards for content created

16
00:00:51.000 --> 00:00:52.700
with the support of generative AI.

17
00:00:53.670 --> 00:00:55.590
On top of it and thankfully,

18
00:00:55.590 --> 00:00:58.860
LLM [Large Language Model] providers are also working hard

19
00:00:58.860 --> 00:01:01.320
to build protections into the models

20
00:01:01.320 --> 00:01:04.170
to mitigate some of the more well-known risks,

21
00:01:04.170 --> 00:01:06.660
such as the copyright infringements,

22
00:01:06.660 --> 00:01:09.360
the lack of truth and biased outputs.

23
00:01:09.360 --> 00:01:13.110
And of course, as these models will get better,

24
00:01:13.110 --> 00:01:16.500
guidelines restricting usage of generative AI

25
00:01:16.500 --> 00:01:21.000
can be pulled back, but it will take time.

26
00:01:21.000 --> 00:01:24.900
CEOs should start with reducing "shadow AI"

27
00:01:24.900 --> 00:01:27.600
by enforcing policies that restrict

28
00:01:27.600 --> 00:01:30.990
and sanction use of tools like ChatGPT.

29
00:01:30.990 --> 00:01:34.380
Then, they should work with their tech teams

30
00:01:34.380 --> 00:01:39.150
to create protections against leaks of sensitive data,

31
00:01:39.150 --> 00:01:42.690
such as sanitizing all data used with generative AI

32
00:01:42.690 --> 00:01:44.790
during these experiments.

33
00:01:44.790 --> 00:01:48.930
And finally, consider setting up a "red team"

34
00:01:48.930 --> 00:01:52.140
to deliberately find failure modes

35
00:01:52.140 --> 00:01:54.300
and vulnerabilities to the business

36
00:01:54.300 --> 00:01:56.130
because of generative AI,

37
00:01:56.130 --> 00:01:59.880
such as unexpected functionality of applications

38
00:01:59.880 --> 00:02:03.153
and enhanced capabilities of fraud and phishing schemes.