Hi, my name is Sven Cattell. I am the founder of the AI Village, which has been around for
seven years, and this will be our sixth EFCOG. This year I'm co-leading it with Rahmat Chowdhury
and Austin Carlson. And it's changed a bunch this year, but the AI Village has been built
up out of the MLsec community and ML security. So we've got a bunch of people who've been
deploying AI to detect malware, detect spam, detect firewalls, all sorts of different places.
And it's quite interesting because we've got a lot of experience deploying AI in adversarial
situations. This year is also quite interesting because it's the 20th anniversary of the first
reported AI incident, which was a Beijing poisoning attack against top file email server.
We have DDS providing some interesting information, a nice website for the inline
people. And we have demo labs that have all sorts of things you can do at the end,
walk to your register. What we're doing new this year, that's different from the usual talks we
give about all the topics that we usually deal with, is we're doing this generative red team.
The red teamers from a lot of the companies have come from AI Village. We have
a lot of the top AI red teamers in the world called AI Village Home, and they have been
playing with a generative bridged machine learning for the last several years, not just the last
year. But the public consciousness with ChatGPT has completely changed the game. We've been
planning for this generative red team since before ChatGPT was released over a year ago.
It's going to be a lot of fun. You get 50 minutes to red team all of the eight models.
You'll see there's some challenges. There is prizes. You can win a 6,000. This is my personal
one. We've got a bunch of new brand new ones. The most exciting thing about it and the follow-up
event is we're bringing security concepts that have been well tested and developed for secure
to make us safer and more secure to ethical issues because the field of AI is really bringing
this, opening up this cannon for ethical problems in machine learning and software. And you really
need this. Security needs to extend itself a little bit to actually deal with these problems.
And we're very excited to be a part of that. After DEF CON, we're going to take all the data
that we received and do a coordinated disclosure process with all the vendors, which is going to
come out in February of 2024. So if you get something interesting, we'll give you credit
and that's going to come out. The thing that we're really excited about is because of all the hype
that we managed to get, we managed to raise a bunch of money to fly in 214 students and seed
AIs coordinating this from community colleges across America. So 214 students getting from AI
and cybersecurity getting full rides to DEF CON. It's going to be a ton of fun. So if you see some
of them, please say hi from me and give them a warm welcome to DEF CON. Thank you very much.
Hope to see you at the village. And if you want to compete,
you should look up the prompt injections and all sorts of stuff. Bye!