112. How To Protect Your Secrets: Inside China’s Technical Surveillance Playbook (Ep 1)

19:55David:Atio even has something called agent collaboration.

19:59Gordon:Yes, but in this case, that means giving people the ability to let AI work seamlessly in the background for them.

20:06David:Try Atio for free at attio.com slash trick.

20:36David:of it can hide what actually matters. Most CRMs capture activity, but rarely turn it into insight. Atio is the next generation CRM built with AI at its core. It reflects how modern teams operate fast, layered, and interconnected. Sync your email and calendar and you're up and running in minutes. From there, teams simply ask Atio what they need to know. It's about deals, customers, or what's moving across the pipeline. Powered by universal context, Atio's intelligence layer, it connects what's happening across your meetings, messages, and data, and it turns into answers you can actually act upon.

21:13David:With Atio, you are seeing what's gaining momentum, what's stalling, and what needs attention next. In an accelerating world, clarity is the real competitive edge. Try Atio for free at attio.com slash empire. The world moves fast. Your workday? Even faster. Pitching products, drafting reports, analyzing data. Microsoft 365 Copilot is your AI assistant for work. Built into Word, Excel, PowerPoint, and other Microsoft 365 apps you use. Helping you quickly write, analyze, create, and summarize. So you can cut through clutter and clear a path to your best work. Learn more at Microsoft.com slash M365 Copilot.

22:04David:Okay, welcome back, everybody. David, I guess it's time we start diving a bit deeper with Glenn into some of the specifics of what this all means for the business.

22:15Gordon:And rather than do this kind of tech first, I think it might be easier to actually just talk about the broader challenge by looking at how it's impacting some important parts of the human intelligence business. So sort of start with pieces of the trade that are being impacted and kind of talk about how the tech is reshaping them. And I think one of them, which is very important, finding the people who have the secrets we want. Typically, you'd call that sort of spotting and assessing potential targets of value, people who have secrets. And that obviously, you know, Gordon, I mean, we talked a little bit in our recent episodes on Chinese spies using LinkedIn, for example, to try to find people.

22:58Gordon:But we've also done it in some of the more historical series as well.

23:02David:Yeah, I guess historically, Glenn, this would have been done by trying to meet people, by trying to talk to people. It's the classic thing we think of, isn't it? Of going to cocktail parties. The diplomat at the cocktail party sizing up his KGB opposite number or similar, trying to work out if he is KGB, trying to work out what kind of personality he is. There may be someone building a file and trying to work out what secrets he has. I always think of Oleg Gordievsky, where they get a tip-off from one person, from one defector saying, this guy's interesting, then they build up a file. And then eventually someone, I think, approaches him on a badminton court, a kind of cold approach on a badminton court in Copenhagen, of all places, to basically say, do you want to talk?

23:44David:I mean, that's the old school way, isn't it? And I guess that's what's got harder. It's harder and it's easier. And let me talk about how it's easier. So we just talked about all this data that's collected on each one of us every single day. I think I read it's 2.5 sextillion bytes of data, which is an absolutely stunning number. Somebody, I think, described it as two Oxford English dictionaries on top of each other covering the entire planet Earth worth of information. every single day about people, right? So that's a lot of data. And if, for example, Gordon were interested in someone, he could use, avail himself of all of that data.

24:29And you don't have to be anybody special to do that. This is not limited to intelligence services. This is what e-tailers and merchants and businesses do all the time to find customers and to detect fraud is they build a profile. And there's books written about this, how the modern surveillance economy determines when you buy, when you die, when you click, who you know, what you care about. And so all that information can be mined. And we've heard about this data mining. And so in that sense, it's much, much easier to find out who would have the kind of information that anybody would be interested in, including a government or an intelligence service.

25:11It's harder in the sense that if you wanted to go meet some specific person and talk to that person to feel out whether, say, if it was Gordon and David that you wanted to find out what David knew and how he was inclined, you're going to leave a record of that. If you call him, you're going to leave a record of that. If you meet him, you're going to leave a record of that. If you see him at a cocktail party and you don't want to do that because if he is valuable, You want that relationship to be clandestine and you've already ruined that from the get-go. You can't put that toothpaste back in the tube.

25:47So you have to be exceptionally careful about making those associations that at some point you would want to hide because they don't go away. And as I've told David in previous discussions, this technology, big data, has really changed previously our understanding and functions of time and place and what you can know about a person. In the old days, if you did an act like talking to somebody at a cocktail party, okay, you had a conversation with someone at a cocktail party. As long as the people there didn't see you at the time and they didn't have an interest in that, that was fine. But the problem today is that meeting lives forever in cameras and microphones and other sensors.

26:32David:The fact that those two phones of you and that person were both at the same place at the same time lives forever in a way it wouldn't have done. That lives forever. In a way with Gordievsky meeting someone at a badminton court, that could have been done quickly and quietly. So, yeah, it's a really interesting point, isn't it? Because we can find out much more about people now. So you could do the research bit pretty easily. And I guess we all know, as you said, this is something we all do. We all can kind of, you can Google people, you can look up social media profiles, so you can get quite a lot of understanding of someone.

27:03David:You could even say spy on them without actually meeting them or having to kind of engage with them face to face. But as soon as you want to do that, engaging face to face or interacting with them or connecting with them, that's where it gets trickier. It's a devilish challenge. It's doable, but it's very challenging. Because if you think about it, you're carrying around this device that's tracking you and you're moving in a sea of other devices that track you. Various ring cameras and traffic cameras and business cameras and other surveillance cameras and CCTVs and microphones and people's personal cameras.

27:42You're in this sea of RF and data collection. and if you try to hide from it, you stand out. And if you don't try to hide from it, you're captured by it. And so you have to find that narrow place where you don't stand out and you're not captured by it or it's not you. So it's exceptionally difficult. It's doable, but it's difficult.

28:09Gordon:One of the areas of this entire kind of UTS environment that maybe we're sort of conceptually aware of, but don't really understand or see most of us day to day is this world of data brokers and commercially available information that's not top secret information in the files of the agency. And it's not open stuff that's on social media, but it's information that if you have money, you can go out and acquire. I'm wondering, Glenn, if you could talk to us a little bit about that space and what's knowable about people if you have a few bucks and know the right data broker. It's a real appropriate question because we tend to think of secrets and open source information as the two categories, but there's other categories and you just identified one.

29:03So there's open source information, which any of us can get for free. We go on our computers or our phones and we go to a browser and we go to Google or Bing or one of the LLMs and we ask, you know, who's David McCloskey? But then there's this entire other category of commercially available information that you have to pay for. There's a guy named Mike Yagley, who's done a lot of work on what data brokers have and how they make their money and what they can know. I spent a lot of time talking to him and learning from him. And it's just amazing. What is collected? And there are these tiers of collection.

29:45And the more money you have, the more information that you can get on people and organizations. And nobody really cares who you are. They just care that you can pay. And I think there's probably tens of thousands of these brokers around the world. I think the United States leads in this because it's the most unregulated or loosely regulated data economy, much less so than, for example, the EU. It's because we have more freedom here, Glenn.

30:18Gordon:I think is what you're trying to say. I'm not going to comment on that, David. It might be more lobbyists from big tech, but the point is anybody can buy this data. It's not restricted in any way, and it's for sale.

30:33David:When we're talking about looking for people who have secrets that an intelligence agency might want, I mean, sometimes people put that on LinkedIn. You know, I used to work for the CIA or I used to work for the NSA or I used to have security clearance, things like that. And also, I guess the other thing you're looking for is someone who had access to secrets or has access to secrets and someone who might have a kind of weakness or a reason why they might be willing to pass on those secrets, like they're in debt or they don't like their employer. And I guess that second category, you can also probably get from some of that commercially sensitive information.

31:08David:If you can do a credit check on someone, you can access their credit, or you can look at what they're trying to buy or where they're going, or is their marriage in trouble? All those little telltale things which might actually be relevant in this situation. Yeah. Well, I think the most salient point that you made is that people share that information without recognizing that they're sharing it. I mean, this is one of the really interesting things about intelligence that people don't realize is that people don't recognize the value of their privacy. And so they undermine it constantly. And so they'll post on Facebook or InstaFace or Insta.

31:46I don't even know them all. Instagram. They'll post, you know, I hate my boss. OK, good to know. All of a sudden, they'll post a family picture and the spouse isn't there or a vacation that they took by themselves and all of these things. So they either say it directly, I hate my boss, or someone who's clever can infer this. There's the famous case of, I think it was Target that sent directed advertising to the daughter in a family for her pregnancy. Well, they inferred her pregnancy from her buying pattern, and she had not told her family that she was pregnant. I think this was a New York Times story many years ago.

32:27And so they either announce this information, I hate my boss, or it can be inferred from the rest of the data that's out there about them.

32:37Gordon:I found a couple fascinating examples that are connected to that, Glenn. I mean, one of them was a former CIA officer after they had left on LinkedIn. LinkedIn suggested to them that they should be connected to someone who had been their asset when they were working at the agency because it had connected, I don't know what country it was in, but the algorithm had said, oh, you've got enough kind of overlapping connections. Interests. You should be connected. I mean, on the point of like even just locational stuff and what you can buy from data brokers. I mean, there was a researcher who had actually done a study for the Defense Department to try to see what might a committed foreign adversary go out and buy on U.S.

33:26Gordon:service members. And so they set up a U.S.-based organization and a Singapore-based organization, contacted U.S. data brokers, and they ended up buying individually identifiable contact, financial, health, and a bunch of other data on thousands of active duty U.S. military service members. and they got it for about 12 cents per service member. And they were able to geofence some of that stuff so that they could look at, you know, sort of where a lot of these people were based, including some sensitive special forces operators. So these data brokers are just selling this stuff to Singapore, selling it to the States.

34:05Gordon:It kind of doesn't matter. And it's really, I mean, that is a great start for a targeting package if you're a foreign intelligence service. And also if you're trying to show the right ads to these service members. Right. So that overlap in sort of targeting effectiveness is is I mean, it's really quite profound and it's astounding what's available if you have some money. Yeah, this is the problem is the data reveal connections that you don't recognize as being revealed. And to your point about U.S. service members, this is a problem that all countries face. There was a story, I think, last week, China gave EVs to top Israeli military officers.

34:47David:And I just thought, no, do not accept them, you know, because a car is like a phone except on steroids, right? And it collects all kinds of information about obviously where you go, but when are you in the car? Who's in the car with you? What do you say? I mean, a car is a giant recording data device. So particularly the EVs, which have all these microphones and cameras, and it doesn't go away. So that was quite a Trojan horse story. So I guess so far we've talked about how data can allow you to understand someone better and how they have. So all of that makes it sound like it's actually easier to find people.

35:30David:Yes. That part of the business arguably has got easier than the days of having to kind of build up gossip over years through kind of meetings and then maybe a cold approach. But let's maybe focus next, shall we, David, on the next stage where it's made it harder, which is actually going to meet someone for the first time, which I guess is about what's called David Cover, isn't it? It's about people taking on different identities and things like that.

35:56Gordon:We talked about this a little bit, Gordon, when we did the series on Matrokin, right? How do you, maybe for the first time, how do you meet somebody and how do you, I guess, frame your employer, your affiliation, your nationality, your name, that piece of it, I think, as we'll talk about, has gotten a lot harder. Because in the Matrokin case, it was relatively, it seems straightforward.

36:25David:When you look back at these cases from the Cold War and even into the 90s, I guess, all these case officers, all these kind of handlers, they just basically had a stack of false passports. It was as simple as that. You had a false passport with a different name in your safe and when you're kind of going on a particular mission you go you know today i'll be james tomorrow i'll be robert pick up the false passport and that was kind of all you needed yes you've got to memorize your backstory and you know where you were born and things like that if someone's asking you but that that was it wasn't it you know for a lot of people you didn't need to do much more to kind of mask who you were and that's certainly what the kind of mi6 team who went out to meet matrokin did they just picked up one of their false identities but then i remember someone from MI6 saying that in the 2000s they suddenly realised that that was getting harder because the ability to maintain your cover when you hit a border they realised as soon as you had kind of Google and social media in the 2000s was getting really difficult because they did a test and they said how long would one of those old-fashioned types of cover just a passport sustain you against a border guard asking questions and with access to Google and social media and the answer was less than a minute, you know, before they realized that there wasn't a backstory, it didn't match.

37:40David:You just couldn't do what you could do in the past. So I guess it was the 2000s, was it? When, you know, people started to realize you just couldn't do this anymore in the same way, travel as an intelligence officer. Yeah. Thank you, Google. I think it really started with Google and AdWords. And again, Google wasn't trying to do this. Google wasn't trying to destroy anonymity and privacy, but AdWords. Which is their advertising business. Their advertising business, which generates probably tens of millions of dollars a day revenue, if not profit. It started essentially this dynamic, and then you reach a point in which storage becomes so inexpensive that it's cheaper to keep most information than it is to delete it.

38:25I think that happened somewhere around 2010. And then, of course, you had the crossover point in which most uses of computation moved over from the desktop or laptop to the telephone. And once it became mobile, well, now you associate identity with location. And that becomes very, very challenging. And I mentioned this colleague of mine, Mike Yagley, would do these presentations where he would show people their pattern. He would buy it from an appropriate data broker. He would show somebody their pattern and they would realize it was actually fairly simple. People do the same thing every day. We're exceptionally habit driven.

39:14We go home, we go to work, we pick our kids up from school. We might go play tennis or go to the gym or something like that. it's pretty simple. Most of us are shockingly boring. And so when we do something that's out of that pattern, that's noteworthy. I think all of us can think of times where you get an alert from your credit card that says, oh, I see that you're in Penn Station, New York. That's not where you live. Is this you? If it isn't, please contact this number. If it is, just ignore it. That's frightening to me. That's a private company that does that. And that keeps that information forever.

39:51And so really, I think this is what you're referring to. This tracking that's part of our normal economic existence affects everybody, including intelligence collectors.

40:03Gordon:And you bring up the fraud point. It's a good one because I guess this same algorithm that is detecting some anomaly in your location or your payment history, it could be used to detect potentially the operational act, right? Because that's by definition going to be a, well, in the past, it would have been an anomaly, I guess. It would have been something that was potentially out of pattern. And again, that fraud detection, I mean, that's like Amex has developed that, right? So that's not being developed by the MSS or the CIA. That's a private company that's got it. I'm curious on the, I mean, one of the pieces of, of, I guess, this UTS environment we haven't talked about in detail is biometrics.

40:52Gordon:Yeah. So the linkage of your retina, your fingerprint, whatever your face with you. Talk a little bit about how that is maybe impacting clandestinity or the work of human intelligence today. You have these sensors which take photos of you from various angles and then process that data about the distances and angles of the various features of your face. And they construct a digital map of your face. And it's your face and only your face and nobody else's face. And it's gotten quite good and it's going to get better. And now it can recognize you when you've got a surgical mass on. Got a lot better during COVID for that reason.

41:31So if you try to disguise yourself, that's a bit of a challenge. So the most obvious one, I think, is facial recognition combined with the statistical automated AI machine learning capabilities and algorithms that make each one of us uniquely detectable based on our face. Then you have iris scans, all of which are becoming more pervasive and cheaper. Of course, fingerprints and people open their phones with their thumbs and they open their funds with their photographs. And of course, this is stored and shared and can be used, again, for law enforcement purposes, for identification, for sales purposes.

42:15Various hotels and retail establishments recognize customers when they walk in the lobby because of facial recognition. I find that unacceptably creepy. But these establishments rely on this and some people like it. Oh, hello, Mr. McCloskey. Welcome to Acme Hotels. Sorry, if there really is an Acme Hotel, I don't mean to disparage you.

42:37David:We'll bleep it out. We'll bleep it out. Yeah. Because I remember hearing that one of the first challenges for the intelligence services was iris recognition. I think certain Middle Eastern airports introduced the idea you were going to get your iris scanned on arrival at the airport. And they suddenly thought, hang on a sec, if I go under that false passport I'd used before and my iris is scanned, if I turn up here next time even on holiday with my family you know and have my iris scanned and it says this is the same iris as a different name they're gonna go you know spy you know and arrest you so that was i guess an early example of how that was a worry disguises in the old days would be like a wig all those kind of things i mean in the movies you'd think there are kind of maybe this is getting into detail you can't get into glenn but in the movies if i'm watching mission impossible or something that.

43:27David:There's kind of fake fingers, fake irises and things like that you can do to fool them. I mean, does disguise still work in the world of biometrics and where they can see how you walk and things like that? I don't know. I think it would be exceptionally challenging. I mean, obviously, if I put on a mask of David's face... Lucky you. You know, I would obviously be much more successful. But if I put on a mask of David's face and I wear a perfect mask of David's face. Sure. I don't know if such a thing exists outside of movies. It's very difficult to disguise yourself. I guess in theory, you could burn off your fingerprints.

44:11But this is all, it's the mathematics of the distance of different parts of your face from other parts of your face, which is highly invariant. Right. And I think also if you If you showed up at a border with a heavy mask, you might get some questions. I wouldn't base my tradecraft on the Mission Impossible movies.

44:32David:No. I seem to remember there was a problem with laughing if you had some of these masks, David, that there was some history where the masks were fine, and as soon as you laughed, you had a problem. So yeah, they definitely had their limits in the Cold War.

44:43Gordon:Yeah. Well, I mean, it also seems like from the standpoint of the disguise, I mean, even if you had one that was quite good, your ability to wear that for any long period of time, I think, would be pretty diminished. I mean, it just seems very technically challenging. I guess, you know, Glenn, the other piece of this we talked about, you know, facial recognition or gait recognition or like person recognition. I mean, there was a story, I think maybe a couple years ago about how the Russians, surprise, surprise, had become quite adept at using the sort of camera networks around Moscow to find draft dodgers, people who were called up to Ukraine and had decided not to go and were still, for some reason, hanging around Moscow.

45:25Gordon:And we're able to make a bunch of arrests around that. I'm curious about some of the back-end pieces of this because we think we go out and we walk around any major metro area. It could be in the States. It could be in London. It could be Moscow. And there's cameras everywhere. those cameras are not always centralized especially in the states it could be a lot of private companies in london i can tell you they are everywhere though right they're everywhere you see them everywhere but there's a back end to this which is the analytics to do something with that information right i'm curious if you could talk a little bit about that sort of back end it's not very sexy but you think about someone's got to take all this data which in in its kind of what's collected is probably pretty unstructured and not exactly useful from an intelligence standpoint.

46:14Gordon:But then the backend, maybe AI powered bit, maybe not to make sense of it is really kind of the beating heart of it. Maybe talk a little bit about that piece of it. Yeah, it's a really good point. The wishful thinking that I get from people when I talk about this is it goes something like this. Well, no one could ever process all that information. They can take all the pictures, but nobody's going to look at all those pictures and nobody's going to find me. I think relying on the incompetence of the adversary is not a particularly successful strategy for achieving your goals. There's all kinds of experiments done with this.

46:56some probably rigged to make them look more successful. But in general, the issue is time and automation. And so, for example, if you talk about China, which is, I think, the most heavily camera-ed up large country in the world, that's a word I've really come to enjoy, camera-ed up.

47:18Gordon:I think there's like a million surveillance cameras in Beijing alone. That's probably it's probably an understatement at this point yeah you look at london which is much smaller than beijing but london i think has tens of thousands of cameras and and beijing has millions yeah um and shanghai has millions um and any large city in in china has millions and uh the the authorities have all the time in the world and all the computational power in the world and all the algorithms that they can avail themselves out to look at these photographs, both video and still private residences and businesses have to share their camera feeds with authorities.

47:59So my advice to people is if you see the camera, assume it works. Assume someone's going to see what's on that camera. Now it might not be today. It might not be tomorrow. It might not be this year. But, for example, if you're committing a crime or if you're doing espionage, an asset caught a year from now is still caught. Right. And so the conversation that intelligence officers have to have now concerns, am I safe for all time? And it's not a question that's easily answered. back in the day before all of this technology existed, an intelligence officer could do an operation, come back home, know that he or she got home safely, find out that the asset later was safe.

48:48And that's that. But now you have to wonder for days and weeks and months and years, if something is going to trigger an investigation that leads a service to go back through that video or still photography. And like you said earlier, I think it was Gordon, connect two people who weren't connected. There could be a newspaper story saying, hey, this person worked for the Russian intelligence service. Oh, I thought he was a businessman. That's interesting to me. I'll go investigate and whether he met with anybody from our government.

49:16David:Yeah. I mean, we've definitely seen examples here in the UK where they've used that kind of CCTV. I mean, the Salisbury poisoning 2018, they were able to identify using CCTV from the train station. And your point about underestimating CCTV and things like that, your peril. I mean, I remember the Israelis, Mossad, normally thought was one of the most kind of capable intelligence services. 2010, they carried out this hit in Dubai against Mabu, I think a Hamas official in his hotel. And they seem to have just not realised how much CCTV there was in the hotel. And there was just all this amazing CCTV of people going into a kind of bathroom, coming out in a different set of clothes.

49:54David:I think a tennis uniform at one point. But, you know, they clearly underestimated the Dubai intelligence, the UAE intelligence service's ability to track them at their peril, because it basically burnt a lot of identities I think they were using. So, yeah, I think, you know, people have learned often, haven't they, from some of those mistakes. So I wonder if we stop there for this time, because I think we've set up some of the problems around these issues of technical surveillance and data. And next time we dive even deeper into some of the kind of ins and outs of running agents and human espionage in the digital age.

50:28Gordon:And of course, if you don't want to wait for that second part, you don't have to. Go to the restisclassified.com, join the Declassified Club, get early access to that episode and to a whole bunch of other goodies as well. But we'll see you next time. See you next time. You