00:00You're also with the White House chief information officer during the George Bush administration. And Teresa I mean an apocalypse
00:05is what we just heard from Margie. Is that what you're thinking.
00:09I mean I think it's interesting. So it could be from the standpoint of isn't it interesting that Anthropic self
00:18-reported that they found thousands of vulnerabilities that could be fixed very easily in sort of the early release of
00:25this tool.
00:26That means that it was outpacing the ability of cyber security products all the big names on the market which
00:32happen to also be now in this cohort.
00:35And so it is interesting from an apocalypse standpoint. It depends on I think your point of view. Is it
00:41apocalypse or do we now finally have a tool that in the hands of cyber security practitioners can make sort
00:48of this limited skill set even more powerful to be able to cover even more than they have in the
00:55past.
00:55I see AI as being an incredible assistant here. The one concern is if we don't have the right governance
01:02and guardrails we could find that this tool or a tool like it could end up in the hands of
01:08the cyber operatives who have negative and criminal intent.
01:11And I think that's the concern here.
01:14That last point is key and it relates to what we've been talking about in the show right the war
01:18in Iran for many weeks we've been saying that Iran as an example has a competence in the domain where
01:24if you're on the defense or you use U.S. based technology for defense that there's every chance that those
01:32going on the offense have access to similar technology.
01:34Although as Margie outlined you know mythos is at the cutting edge in terms of models. How do you see
01:42the field in that in that conflict offense versus defense.
01:46You know well first of all even though there is thankfully a ceasefire has been announced that does not mean
01:52that the Iranian cyber operatives have taken their fingers off keyboards.
01:56So we need to continue to have a heightened sense of vigilance and alertness around these issues.
02:03But what I will say is you know that they are obviously leveraging AI and different types of tools to
02:10be able to carry off attacks like they did against striker and some of the other attacks that are ongoing
02:15against critical infrastructure in the United States.
02:18And a lot of bulletins have come out if people are listening to the show right now and you're thinking
02:22I don't feel like I'm in the know in this information.
02:24I certainly can't give it to you publicly but you can contact your local FBI office and they will tell
02:30you all the things that you need to look for as it relates to Iran and some of the cyber
02:34attacks on critical infrastructure.
02:36What I would say is you know I would love to hear more from Anthropic once they have this cohort
02:43do sort of an alpha pilot I guess is the best way to put it.
02:46What is the beta pilot and what is the you know sort of a rollout to trusted vetted you know
02:51smaller maybe privately held firms look like to make sure that all offensive and defensive teams have these capabilities and
02:59are able to use them against the attackers.
03:01Time is of the essence we're in a race and we want to make sure that everybody who is on
03:06the ethical side of cyber security is actually empowered engaged and has all the best state-of-the-art tooling
03:14possible to fight back these attacks.
03:17Teresa you make a good point and it's actually we've been hearing from Anthropic as they've outlined what glass wing
03:23really looks like and you can see what certainly Newton Cheng who's Anthropics Frontier Red Team Cyber League has been
03:28talking about how this is an industry-wide problem.
03:31You know for the smaller companies that would like to be able to help there is obviously a chunk of
03:36change being offered by Anthropic as well in terms of computing tokens.
03:39How much have you been working perhaps with OpenAI with its well prototype it's been letting and working with others
03:46in the security space to try and ensure that defense has its tools ahead of others as well.
03:51Yeah I think there's there's a lot going on in this industry and so I have been talking to all
03:57of the different sort of leading AI providers to find out
04:00what are the best offensive defensive tools what are some things that are yet to be built what is on
04:06the roadmap and then what are really the best solutions depending on the size company you are how much money
04:12you have because you're right.
04:13Everything is token based how can companies learn if they don't have a big technology team how can they learn
04:20to optimize the use of the tools so that they're spending their tokens sparingly and get the most value out
04:27of the money that they're spent.
04:28And so there's a definitely an arms race you know a healthy competition across the different AI companies and some
04:35of them are now cooperating as part of this Anthropic rollout.
04:38The other thing I do want to mention is that Anthropic did actually self-report that Mythos went rogue.
04:45And so I think that's something also that we need to you know be discussing is that although it's going
04:50to help it also has some fine tuning that needs to happen.
04:54OK
Comments