Ai has been embedded in our society and our technologies for decades.

The AI of last year is not the AI of today, and it will not be the AI six months from now that we think about when we think about national security challenges.

This is an evolution in technologies, and we need to harness for national security.

Welcome to another week of No Such podcast. I'm one of your hosts, Cam Pods, alongside my co-host.

I am Brian Fassler.

Yes, and we're here to talk about the evolutionary technology, artificial intelligence. That's right, AI. What is it? What does it mean to national security? But more specifically, what does it mean to the National Security Agency? Well, today we are joined at the table by two of NSA's subject matter experts Ms. Tahira Mammon of NSA's AI Security Center, and Mr. Vin Wend, NSA's Chief Responsible AI Officer. Thank you and welcome to No Such podcast.

Thank you. Thank you.

All right, so we are here at the table to talk about AI. Vin, could you take us down the road of what is AI? Is it here to stay?

Well, AI is definitely here to stay. We've been using AI, and you all have been augmented using AI that you don't even know. Really? You are using AI for a decade or more. But AI have been sprinkling in the background, and it's technical. People don't pay attention to them. But since last year, we have generative AI and various chat bots coming out, and now it's in your face, and now you're like, Oh, my God, AI is here. But AI has been embedded in our society and our technologies for decades. But when you really think about AI, and the way we think about this, is that really it's a frontier of automation. We're quite comfortable with automation. We have done this since the 18th century. We automate as we go. And now we, as a human society, we have this opportunity to automate a very special set of activities. And this is quite unique, is automating the way we see, the way we hear, the way we comprehend languages, the way that we reason. Now, that sounds scary, but your researcher have been working on this for a long time, and we have been embedding AI in our facet of our lives already.

So I would say this is really about automation. And And then, USA has been a leader in integrating these technologies into our mission. And so there's nothing to fear. This is an evolution in technologies, and we need to harness for national security.

Since you run since 18th century. That means that it's probably changed a lot over time, right? I mean, I probably take advantage of AI every morning when I want to know what the weather is, when I want to know how to get to work. So tell me what traffic to avoid, things like that. I think to my understanding, one of the best principles is it's only as good as the data that is captured. Can you talk a little bit about the data that's being used and the importance of making sure that it's all-encompassing and what happens if you don't have the accurate data in your data set?

At the core of the technology, it's really about quality data. And so, yes, we all have algorithms, we have compute power. But at the end of the day, the power to bring AI machine learning to bear for our organization is the quality of the data. And so we cannot just use any AI. We have to make sure that we have the right data that is legal and compliance and privacy-preserving to really deliver on what exactly we want to achieve. And so data is really key. And I think a lot of time people don't talk a lot about it outside of some AI application. But here at the National Security Agency, we We really care a lot about the quality of the data, how do we harness it, and how do we protect and secure it.

Yeah, I would say from the cybersecurity perspective, that the quality of the data has been saying is so important, but also making sure that we maintain a focus on protecting it because your data is only as good as the security mechanisms that you have around it, right?

I think that it's great to hear compliant and safe when it comes to implementing understanding this technology and having this as an available thing and something that the agency is in the forefront of leading.

Oh, absolutely. Ai is actually not new to the agency. For those who don't know and for the audience, you may not know, but the National Security Agency had been working on AI and machine learning since AI started, back in the '60s. So this is a very mathematical go, like deep math work that NSA had been involved with. So it's not new to us. We have a long history of doing this. We understand the potential, but we also understand the risk. And so we are taking a very set of measure steps to really do this right and do it in a responsible way.

So we can't be the National Security Agency and not talk about the National Security security challenges that AI may pose. So Tahira, could you outline what some of those challenges are?

Sure. I would say there's two things that we're really looking at. First is the pace, the pace of capability development that we see. The AI of last year is not the AI of today, and it will not be the AI six months from now that we think about when we think about national security challenges. And the second is the scale, the scale of implementation, the availability of AI in the public, and then specifically with our foreign adversaries, the way our foreign adversaries are advancing their capabilities with pace, the way our foreign adversaries are scaling, implementing, and what that means for our national security, and especially this year in an election year, election security and the role that AI has to play there is a key focus for the national security agency.

Yeah. I think it's a challenging time in term of we see China as a pacing threat, as a Rishy competitor. And when you look at the PRC or the People's Republic of China, and the government, they're very focused on investing in AI to gain that economic, diplomatic, political, and military advantage. And so they see AI as a core technology and a core opportunity for China to grow and out-compete and out-maneuver, outmatch the United States and the West. So it's It's quite challenging. We do not want to live in a society where our AI are built, run by the PRC because they are not built on our democratic values. And so So we have to think strategically on the natural secure implication of that level of strategic competition. Now, at a more obvious level is that, like Tuna mentioned, is that are investing, using, exploiting the technologies so they can embed them into military weapons, to this information campaigns, to cyber operations. My assessment is that the national security threats that we're facing are just being amplified by AI. When you see not only nation states, but also nonstate actors are using AI, including cyber criminals, to run their campaign.

So we are seeing the amplification in term of the scope and maybe the consequence of the threat. And so we have to really understand how this threat is manifesting and also how to manage the risk, and especially how can we manage the risk of foreign cyber attacks against our own mission capabilities, national security system and weapon systems. We need to secure the system to maintain that advantage.

I think the proliferation of generative AI capabilities, just out and available to the public We've already seen how that has an effect on non-nation-state actors being able to leverage it in ransomware, in cybercriminal activity. Most of us have probably heard in the news about different audio, video, synthetic media generated to sound like somebody you know or somebody you trust or someone in leadership that helps cyber criminals reach their usually financial aims, but not always. Disinformation campaigns can have wide-reaching impacts, and AI presents capabilities which have great power for good. But in the cybersecurity landscape, we also have to be looking at how they can be used for for deep fakes, right?

In terms of the disinformation, it's an election year. Nsa has had a role in ensuring that our elections are secure. Can you talk a little bit about What are we doing in 2024?

Absolutely. This year, people are talking a lot about AI because we know that our adversaries, state and nonstate, are leveraging generative AI to influence or aiming to influence our election. Nsa, working with our US government partners, are really trying to deliver our exquisite intelligence insight that we can get to secure our election to work with the partners so that we can bolster countermeasures and ensuring that our elections and our infrastructure is safe and secure for the American public.

Seems like the The power of partnerships is the superpower here. Without it, we're not able to be effective, but with it, we're even better. Okay, so in the topic of election security, Tahira, you mentioned the pace. You talked about the scale. With the AI Security Center, former Dernza, General Paul M. Nacca-soni, stood up, USA's AI Security Center in September of 2023. It's now stood up. It's now in forward motion. Now, this center, the center that you lead, what is the AI Security Center doing to keep our country safe.

The AI Security Center keeping in mind the responsibility to secure national security systems in the defense industrial base and to enable them to be using cutting edge technology in a secure way, because that's really what we have to do for all of national security systems, we'll be accomplishing their mission. Ai is a critical component of the technology that they need. And then I say, and the AI Security Center, we aim to provide the security, the best practices that they need so that when they're implemented targeting AI, they know that they're doing it with security at the forefront. We have just stood up, hit IDOC at the end of December, initial operating capability. And what we're really trying to do is get information to our customers in the national security systems, in the defense industrial base, about how to secure their AI as they're implementing it. I think if you think about the history of the Internet, for example, we make the Internet and 10 years later go, Hey, we should encrypt some stuff. Or, Hey, how about certificates? And so AI is rapidly expanding, rapidly growing just within society, right? And security, we know better now, right?

We know that security needs to be at the beginning, at the implementation phase of that. And so we in the center, seek to understand the threats given the exquisite insight and intelligence that NSA has across its two missions, SIGINT and cybersecurity, to take that information and connect it to researchers that work within our center who are brilliant. And now we're doing that work to take those insights, and when we can get them to the right classification level to have engaging information exchange with industry who's creating, running, right, delivering these solutions to the national security systems. Because that's how we inform the researchers on what the threats are, what the capability is, and what we need to do to secure it. When we get all those pieces lined up, then the delivery is best practices. Nsi says, This is the best practice for deploying your AI. This is the best practice for your model security. This is the best practice for access management and authentication. We're getting our feet under us right now and then what the year looks like for publication.

Awesome. You talk about mission and how potentially AI can be used in that area. Can we potentially get into maybe an example of how AI was used in a mission success here at the agency?

Yes, absolutely. I think it's not just foreign intelligence, but using in cybersecurity as well. For those who may, may not know, and as I say, has two major missions. One is to deliver foreign intelligence, true signos intelligence, and the other one is cybersecurity. For foreign intelligence, we have been leveraging AI, machine learning for decades, long, long, long time. Just imagine the work that our analyst do. How can they understand a foreign leader's speaking, their planning, and what foreign militaries are up to? We all communicate. That's the point of signal with intelligence, meaning that we also have to understand what we hear, what we see, and also how we comprehend language. As you can see, AI allows us to do that through machine translation. What we found was that AI can be embedded through the entire intelligence analyst workflow, and then we insert AI to augment every step of the way. We've done a lot of this work, and we don't shy away from it.

Can you talk about the authorities that are in place to ensure that we're using AI responsibly?

I think one of the most important responsibilities that we have as a cybersecurity mission in NSA is helping to protect the national security systems and the defense industrial base. The director is the national manager for those systems. And so in the cybersecurity directorate, which is where the AI Security Center is located, that is a primary focus for us. And so national security systems are weapons and space platforms and any network that has classified information on it. So our nation's top secret to reside on these networks. And in the cybersecurity mission, we seek to provide the cybersecurity for those. As technology advances, and AI becomes a part of what NSS owners or the defense industrial base wants to implement into their networks, then security of those AI systems becomes really important for just comprehensively working towards our shared security.

Yeah, I want to add what Tehira mentioned is we have a lot of discussions in terms of protecting privacy, Security, safety, and whatnot. But at the end of the day, AI security is really fundamental and foundational to protect everything above. A lot of times people will say, Oh, AI safety, AI safety. That's okay. But if you cannot secure your safeguards, you don't have any safety. A good example I can think about would be like a stop sign. Or if it's more dynamic is a traffic light. A traffic light is a safeguard to ensure that we drive safely and prevent accidents. But if you think about it, if the traffic light, it secured and hackers can hack the safeguard, then you actually have no safeguard. You have a beautiful traffic light. Nobody obays.

Traffic everywhere.

Traffic everywhere. You have the traffic light, it doesn't do you any good. So I think that is how we think about security. It's so foundational when the nation talks about safety, and we recognize the many opportunities and challenges of AI safety and the potential of a harm that could cause to user by AI. But we want to also understand that in order to have the safeguards, we also have to secure them as well. And so we're working with literally everyone trying to figure out how to help the nation on that front.

Awesome. Now, I know we can talk about artificial intelligence. We can talk about AI all day. And even while we're talking right now, during this podcast, the technology itself is advancing. To actually wrap this up, I would like to go into talking about AI in the future, where it's headed, and what do we see. Is society prepared for where AI is headed as a technology? Vin, we'll start with you.

Yeah, absolutely. I would say that I cannot really tell you what fancy technologies will be popping up, but what I can tell you is that NSA is watching this space very closely to understand the opportunities, but also manage those risks that we see, especially AI security.

When you say, Are we ready? That is the mission of the AI Security Center. It's for national security systems to be ready and secure as they implement AI technologies. So we're doing all that we can at NSA and within the AI Security Center, investing our resources, our people, our expertise, and bringing what we have to bear as an assay to the problem.

Well, thank you so much for sharing that. I must say, while we may not necessarily know for sure what will happen, at least we know that we have professionals in our NSA, subject matter experts working in partnership with other industry and other experts to just counter the measures that come with this technology and where it's headed and what's going to take place. So if someone wants to work within the mission of AI at the National Security Agency, where should they go to actually see about those opportunities?

Yes, go nsa. Gov, and we have a series of AI jobs that you can apply to, AI engineers, data engineers, MLOps engineer, data scientists. Feel free to link up, and our team here will be in touch with you.

Yes. And I would say that, and I say has a really robust program of internships for college students and even high school interns who's on the cutting edge of technology, usually college people, right? Yes. Those applications usually run in the fall, but all available on nsa. Gov. A really important way that we bring talent into the agency and new ideas, right? Because you have to have an innovative mindset to be doing AI security work, and we are interested in those people.

Well, high school students, college students, you have heard it here first. The opportunities are available for you for internships, and of course, for anyone else, you can head over to intelligencecouriers. Gov/nsa. Is there anything else you all would like to share with us today?

I would add that anyone that's interested in the work that happens at NSA, cybersecurity is the place where you can learn a lot about what we're doing to defend our nation. So on nsa. Gov/cybersecurity, you can find the cybersecurity products that we put out about what the adversary threat looks like, how to secure your network, how to do zero trust, how to do cloud security. There's a ton of great content there, including some AI Security Center publication about secure and resilient deployment of AI systems. So recommend everybody check us out.

All right. Well, Tahira, then, thank you so much for joining us today on No Such podcast. It's been a pleasure to have you here with us to talk about artificial intelligence, where it's headed, and what the agency is doing to get at on the forefront of this technology.

Thank you for having us. Thank you.

All right. Thank you for joining us. It's been a pleasure being your host for this episode. My name is Cam Potts alongside my co-host. Brian Fassler. All right. And we look forward to seeing you next time on No Such podcast.

Thanks for listening to this episode of No Such podcast from the National Security Agency. If you enjoyed the show, please leave us a review and make sure you're subscribed so you don't miss our next episode. For show transcripts and other information, please visit nsa. Gov/podcast.