Hi and welcome to today's episode of the AI Futures podcast series, where we're bringing you some of the world's best thinkers and leaders from the world of AI. Today I'm delighted to have with us Anne Neuberger, who served as the Deputy National Security Adviser for Cyber under the Biden administration. We're happy to have her here at the summit, and I'm excited to be in conversation with you. Anne, welcome.
Thank you so much. It's fantastic to be here. You feel the energy in the air, so I'm glad to have this conversation.
Yeah. No, I think the summit has been—it's been a huge summit. I think we've had participation of hundreds of thousands of people, and I think the energy is palpable. The excitement about AI amongst, it seems, students, policymakers, corporate leaders is evident in the hallways. But what I wanted to do with you was talk about how AI is intersecting with cyber. You've of course served in the former Biden administration. Talk to us about how you are seeing the cyber landscape first evolve around the world.
Absolutely. So first, what we've seen is, from a geopolitics or national security perspective, geopolitics play out in cyberspace. So we see the core infrastructure that our economies and our national security rides on has, in many cases, been vulnerable to cyber attacks, either by countries that want to put pressure on another country or by criminals who sometimes just want to scam an individual or get a company to pay a ransom. And we often say that it's far easier to attack in cyberspace than to defend. An attacker has to find one way in; a defender has to watch every door and window. So certainly AI helps both. AI helps attackers in terms of, for example, accelerating finding a vulnerability and building an exploit, and AI helps defenders. And I actually think it will help defenders more, and that's the space I'm the most excited about.
Yeah. The defenders, of course, have maybe had a slower start than the attackers, as you're mentioning. Is your sense that this is going to be a continuous cat and mouse game? I was talking to Jay Heaps, who heads a firm called Zcaler, and he was saying it's a race. They're going to run fast; we have to run faster. Is your sense now this is just going to be a continuous race to make sure that you're always a step ahead as the defender?
You know, quite frankly, I've worked both on offense and defense, and I think the race is now totally changed. It used to be human against human, and how quickly humans could go through the logs, find something that looked abnormal, determine if it was malicious, and then action it. And like you said, defense just took longer. Today, it's machine against machine, and the ability of AI to help identify what's normal in a network, to help identify an odd pattern, and then action it. Right now, it's often a matter of speed, right? Attackers steal passwords. That account wasn't set up with multifactor authentication. They use the password, get a foothold on the network, try to scout around and find something meaningful, and then the hunt's off, and then it may alert a human, and the human looks into it. Is it a false positive? Instead, with AI, you know, when you have an account that's logging in at an odd time doing something unusual, we can get to a place where an agent can action it. We're not there yet, and there's a lot of security we have to build in so the agents don't become an insider threat.
But we can see we're in the machine against machine battle. Systems can make a big difference.
Yeah. Let's also talk about the threat actors here, right? So India—at least in the context of India—the traditional threat for India from a national security standpoint, the traditional one has been from terrorists. Terrorists across the border, terrorists in different parts of the country, etc. And those terrorist groups were groups that, you know, the national security establishment would start to track, cover, understand their modus operandi, etc. It seems in the world of cyber it might be different, might be similar. So talk to us about these threat actor groups, how quickly they disband, are they more state, more non-state, and how are states now thinking about working with these groups compared to, let's say, the global war on terrorism and how, you know, both countries like the US and India handle that threat?
It's such a good question because what's so different, to your point, from the global war on terrorism with what we see today is still in the physical space. Countries have borders, and as a result there are traditional laws and policies around sovereignty and how you defend a border. Digital space is fundamentally borderless. So, as a result, you have a mix of actors, and it's sometimes hard to tell. Is it a country conveniently looking away as a set of actors, activists or criminals conduct attacks from within their borders, or was it a wink and a nod? Go ahead. And when you have infrastructure, different countries have different approaches. China has state-owned enterprises, so the Chinese government often monitors water or power systems. In the United States, those are owned and operated by companies, and they fundamentally make cybersecurity decisions based on cost and revenue. But yet, this is national infrastructure that citizens rely on. So to your question, there are three categories of actors, but it's really gray in terms of where they fall, and those are countries, you know, transnational groups like drug traffickers and terrorists, and then criminals out to make a buck. And that's where we've seen ransomware in the United States, for example, shutting down hospitals because they know that companies will pay. So as we look across those three categories, AI accelerates all three. It also, you know, as part of trying to counter that kind of threat activity, the first step is to determine who is it and what are the rules of the road for them. So if it's a country, there's actually rules of the road countries agree to. Now what you do when those are broken is what's challenging in cyberspace. But right now, that's where this is playing out.
Yeah. And you know, for a long time people have been talking about cyber. I've been in the tech world now for over a couple—almost a couple of decades. A lot of people have been talking about the threat of cyber for a long time, but it feels like, with the exception of maybe a couple of countries, many countries have not taken it as seriously as they should have. And when I talk to others in the cyber world, they say that unless there's a massive incident, countries will not stand up and take note. I remember Colonial Pipeline happened in the US a few years ago. Do you think it's incidents like those that have really brought this to the fore, and will that be the case for other countries also? Did something fundamentally change in the US after the Colonial Pipeline incident?
It's such a good analysis. You know, when cyber attacks are invisible, it's hard for people to feel it. So even if massive amounts of data is stolen, it's hard to see it. Unlike a terrorist attack where suddenly you see a bomb go off or a car being blown up, you see it very visually, and of course that shakes you up maybe more. Exactly. It really does. So what we see in cyberspace is two things. One, when it plays out in physical space, people pull up at the gas station and they can't get gas. People check in at a hospital, the hospital's network is encrypted. So unless it's a critical surgery and then they'll have to hire runners to run images into surgery, they can't provide services. So then if people start to understand, well, the digital world actually affects me. Yeah. And I think as well, you know, during the war between Iran and Israel, Iran would compromise plain old street cameras in cities in Israel to assess how many buildings got hit by missiles. Was the drone attack successful? Did it hit a home or did it hit an office building or did it hit something military? So we don't often think about that piece as well. The very tech on the street or in your home can also be a source of national espionage or national pressure as well.
And as a result, the other piece that I feel is happening that I wanted to ask you about is war has traditionally been on the borders of countries. Citizens and the common people have typically not been under active attack. Whereas in the world of cyber, it seems, as you're describing, whether it's hospitals, whether it's street cameras, or whether it's my phone, it seems like that in some ways in this new state of cyber warfare that our world has entered, everyone has a worry.
Yeah, to your point, you know, Raj Shukla and I did a talk on Wednesday here about that, and I noted that we see three things have really changed in war. One is definitely first just the commercialization of defense technology. When you look at drones, those are often built from commercial parts, weaponized. And you look at a country like Ukraine that lacked a navy and used drone swarms to take out the Russian Navy, right? So that commercial tech being used in war lowers the barrier of countries engaging aggressively. And to your point, whether it's countries or terrorists or whoever it is. The second piece is, as a result, the cost of defense far exceeds the cost of offense. So the United States engaged when the Houthis tried to shut down maritime traffic in the Red Sea. The United States worked to try to counter Houthi drone attacks against ships. We countered it with missiles that often cost much, much more—orders of magnitude. That's not sustainable over a period of time, right? You can do drone swarms versus when each missile—and they're not designed for that. So what keeps me up at night is actually how we defend, to your point, in a borderless world against drone swarms if it's over a game, right? The cost of that defense and how we think about electronic warfare and making that more commercialized, and so that we can actually defend against this, is key. And then the third part I would say is the role of AI in war, right? So we've seen that evolve. In some ways it can make potentially war more precise because we struggle sometimes with terror groups hiding among civilians. You know, how do you identify which is which? And by patterns of behavior or communication, potentially that can get better. On the flip side, AI can help a set of drones communicate or, based on geographic assessment, determine where they're going. So they don't need a connection back home, which is typically the way we disable them from a defensive perspective. So those three factors coming together have really changed war across the commercialization.
Yeah. But how do you start to protect citizens, right? So for a country like India where we have a billion plus people, everyone has devices today. The US has over 300 million people with devices. Again, if citizens start to become the targets through the devices they are in, is that something that you feel policymakers are spending enough time on? Everyone seems to be talking about the systems and the infrastructure, the critical infrastructure, but are we talking enough about the regular citizens, regular phone-carrying citizens who could also become the carriers of some of these attacks?
Not enough, because what we essentially have is, from a national perspective, we want to ensure that a citizen knows that when they turn on their light switch or they turn on their water, the light will go on and the water will be clean. If it's in a—that those are increasingly digital systems. So the United States, for example, we had a good collaboration with India, bringing our experts on energy systems to work with Indian experts to think about how do you secure an energy system, particularly energy systems that are getting more inputs from a person who puts a solar panel on their roof, right? Every time there's an entry point, that's an entry point for good and for bad. And in many ways, that's a good example of, you know, what's expected of a solar panel. So you can really whitelist and narrow and only allow for that. But sometimes these systems are not designed in that way from the beginning. That's one thing that I think is exciting about AI, in that you're seeing now from Google, from OpenAI, from a number of companies, the ability to look at a set of code and automatically find vulnerabilities and patch them. So I'm hopeful we'll have more secure code generation, which can help us in that case.
Yeah. Let's also talk about the private and public cooperation here, right? So I think one of the key things in, again, cyber is that maybe the biggest capabilities lie in the private sector. Biggest capabilities possibly for defense, but also possibly for offense, especially countries that have large tech firms like the US does, now China does, many other countries do. I remember that there was a lot of attempt after Colonial Pipeline, but also some of the other attacks in the US and elsewhere, that we've got to have the private sector be a lot more transparent about when something has happened, right, and share that information with the government. That seems to be now a standard practice. But in your experience heading up cyber in the US, what were some of the lessons you've learned that could apply to countries like India and elsewhere about how to ensure that the private and the public sectors here are actually cooperating on the cyber front?
