Is your boss watching you? The rise of AI surveillance at work

This is the Guardian. As of 2025, a third of UK companies were using so-called "bossware" to track workers' activity. And to be fair, that was probably an underestimation. AI is only making this kind of surveillance more sophisticated. And it could be coming to your workplace. So today, the rise of AI employee monitoring, and what it does to those being watched. From the Guardian, I'm Ian Sample, and this is Science Weekly. Robert Booth, you're the Guardian's UK technology editor, and you've been reporting on the increase in AI-assisted monitoring at work. Give me a sense of the kinds of things this could be. So this is recording calls, looking at email transcripts, checking your mouse movements, sometimes checking your keystrokes, seeing what kind of files you've opened, what you might have done with them. It can go as far, in some cases, as even using kind of cameras to see how people are moving around the offices. But lots of the time, it's kind of recording of phone calls or recording of interactions, and then using the AI to produce instant transcripts and then analysing those transcripts to work out what's going right, what's going wrong. And monitoring performance isn't anything new or necessarily bad. I mean, people might be used to hearing that a call with a customer services agent is being monitored for training or quality purposes. So where does AI come in? How does that step things up a gear? Yeah, well, I think it's the sheer capability of a technology at crunching through huge amounts of data incredibly quickly. And it can never be sated, kind of, with AI. The more data you give it, the happier it seems to be. So what you have is a difference between kind of the occasional check, which is kind of all that a manager could previously do. So having constant monitoring, you first reported on this last month in relation to a company called Multiverse. They supply AI and digital skills training to workers in the NHS, local councils, universities, the private sector, including here at the Guardian, I should say. Tell me a bit more about Multiverse. Tell me about them. What do they do? So this company is set up by you and Blair, the eldest son of the former British Prime Minister Tony Blair. He's kind of envisaged it as a kind of fast growing tech company. Essentially, it's training workers and apprentices in how to use technology. He's really got switched on to AI. He talks about kind of employers wanting to be, he says, as AI piled as possible. So he's a real kind of advocate for AI, AI piled, AI piled, yeah. Well, I think that, I think that essentially means is you're a believer. Right. You're high on AI. You've, well, I'm not about high, but definitely a believer, possibly high. He calls it an upskilling platform for AI so that the companies and, you know, as you say, the public sector, the employees are starting to use AI better. But he's also using AI to monitor the teachers that he has. They do online classroom sessions and they've started last month to use an AI system to monitor every aspect of those interactions that they're kind of teachers and coaches are doing. And so what are the whistleblowers there being saying about what's been going on? Well, like I say, this started last month so it felt kind of very new to them. It's a real change in their kind of working circumstances. So for example, they were saying that, you know, the AI will pick up on how long it takes them to fix glitches with the IT. If the kind of call drops out, if they haven't got it clicked on again within a minute or so, then it'll kind of flag that or if they're using too many filler phrases like sort of and kind of things that I've got to watch out for as we do this, then they get flagged for that as well. And then the other thing which I think is kind of really, really kind of, he says as he fills some filler phrases, is that when the students go quiet or they only respond with kind of single word answers, that's also seen as a kind of a bad sign. And equally, if a teacher gets pulled into a one on one exchange with one student for too long, that can be seen as a sign that the rest of a group is being sidelined. It is going into really fine detail and it's the AI that's kind of analyzing all of these things. And you can see how it could do that in a way that a human couldn't at scale. And then it provides a kind of percentage score, a percentage confidence score about how they're doing in these different areas and a kind of risk status so that the senior people at Multiverse can kind of look at the teacher's work and give them feedback and the managers can then talk to them about what they want to do differently. And importantly, I think, I mean, what Multiverse makes clear is that the AI isn't replacing the human judgment in terms of how the teacher is performing. Rather, it kind of directs the manager to what is the area that we should be focusing on with this teacher, you know, by highlighting the problems quickly and that it's essentially using the AI. And I'm quoting here this spokesperson using an AI to improve the experience of learners and customers and that in the end, only human managers will be the ones that write performance reviews of the teachers. Rob, do we know if some sectors are doing this more than others or some embracing it? Yeah, well, I think one of the quite interesting things in looking at this is realizing that, you know, this kind of monitoring has been going on for, you know, hundreds of years in industry, in factories, you know, where the outputs are a bit more binary, aren't they? You know, they're a bit more obvious. You can see how many widgets are coming out of the machine at the end of the day, how many man hours of clocks in and out. But this is allowing it the same sort of surveillance, if you like, to move into the more amorphous world of white collar work. And so you're seeing a lot in kind of call center operations. We're seeing in finance, in banks, for example, training companies like multiverse and also kind of AI companies themselves, like tech companies that were obviously kind of more switched on to it. That's, there was the case of Meta were using key stroke monitoring of their workers and, I mean, it's very controversial and there was pushback and I think they paused it. But that was because they wanted to monitor what their workers were doing in order that they could use that information to train AIs in order to do those kind of jobs. So there's different reasons for doing it. Sometimes it's quality control, but sometimes it's also training AIs to replace these people. Monitoring them to replace them. I was going to ask you about that is what are the companies saying are there reasons for doing this? Well, I think one of the things to bear in mind about this phenomenon is that it kind of coincides with the pandemic and the rise in working from home. So there was a kind of an obvious incentive there for people, and a need for people to be more connected and so monitoring starts to come in more easily at that level. So one of the things they say is this is enabling people to work more remotely because we can keep an eye on you more remotely by using AI. You don't have to be in the office all of the time. I think also there's just a straightforward answer which is that it's obviously going to be more efficient to use an AI if it's effective. That is the human monitoring managers don't really want to be spending their time doing the monitoring. They want to be spending their time kind of solving the problems that the monitoring throws up. So that's obviously a good thing for them. And they're also saying that it kind of promotes the development of the worker because if they are taking the feedback and they're getting better at their job. And some workers do talk about this themselves as well, which is actually oh my god, I didn't realize I was doing that. The other thing multiverse said was that the agent, the AI agent, provides instructors with improved advice on development and feedback. And they said that there's exactly how any learning organization should actually be deploying AI. And then they said the most consequential performance management action it can take is to recommend the human personally review session. There's a really interesting question about sort of transparency that comes into all of this. And companies who do this, so they're obliged to tell their workers that they're doing it, when they're doing it, what they're doing exactly, or if not how a worker's finding out. Well, yes, I mean in the UK they have to, employees must explain the amount of monitoring they're doing quite clearly. They're supposed to do it either like in your contract or in the staff handbook. So it's known about, I mean in the cases that I've been writing about the last few weeks, you know, the employees have been told that this is what's happening. Their issue is how it actually affects them. And there are kind of rules around it. It's got to be necessary, it's got to be proportionate. You can't just egregiously do it. There has to be a purpose behind a legitimate purpose behind it. In the EU, the rules are a little tougher. They're allowed to do productivity scoring with this kind of surveillance. But things they can't do under the EU AI Act. They're not allowed to use the AI monitoring to recognize your emotions and your emotional state and to kind of try and detect demeanor. And also they're not allowed to kind of socially score you. So they're not allowed to kind of rate the nature of your education. interactions, how often you might be taking breaks. Things that aren't necessarily like core related to your job performance, those things are prohibited. Coming up, the psychological impact of workplace surveillance. You said Brexit will be a galvanizing. Deliberating, empowering moment of patriotic renewal. Yes. So, are you feeling galvanized, liberated, empowered? Yes. All of them. Yes. All at the same time. You're lucky, lucky man. For almost four decades, Michael Gove has been one of the biggest brains in conservative politics among the rights leading thinkers. Confidence in his shared British identity has taken a bit of a battle. From hugging hoodies to dreamy up Brexit, from Margaret Thatcher to Kenny Badenock, he's both led the right and seen where it's going next. You do identify politics, except the identity that you're attacking is Muslim. No. What we're attacking is Islamism and political Islam. Around the world, the right is firmly on the rise. So, what does Gove and the rest of the right really want? This is Radical Thinking, a new series from the Guardian with me, Edit to your Chakraborty. I'm out Saturday in your politics weekly feeds. Rob, you've course spoke to some of the employees of multiverse. What did they say about the impact that this was having on them? They've talked to me about kind of the feeling of it being kind of a remorse of the system and on the stress that comes from it. And the fact that some of them are losing sleep, worrying about the scores that are getting, say you're doing six, seven hours of condensed work, put some breaks around that, and you're being monitored for six to seven hours. It's exhausting. I've noticed that there's some new lingo that started to emerge around it. So, for example, there was one example of a boss who'd gone to their worker and said, "I've heard some signals about you." And they were like, "What do you mean signals?" And that was signals from the AI about the way that you were behaving, which was a kind of like, that's a strange thing. This is new office speak. New office speak. Yeah, I've heard some signals about you. Right. It's not gossip. It's signals, it's what the AI has been telling me. Even more mysterious and therefore more worrying. It's a bit odd. Don't even know what that means. I did speak to one person who's been seeking therapy about it because it was causing them such anxiety that they were very worried about their performance levels. I'll quote someone, "You're being watched. It's different from occasionally being listened to for compliance purposes. You're being monitored absolutely every word that you say and every time that you utter something. It just feels really dystopian." It does sound like a pretty miserable environment to be working in. Has there been any kind of scientific work on this yet? Or was this just all way too early for science to have had a say yet? It's quite early, but there have been some papers done over the last two or three years. There is a kind of associated, positive association between surveillance and what are called secondary work stresses. So under surveillance, you're more likely to feel under job pressure. You're more likely to feel a reduced sense of autonomy. And you're more likely to feel that your privacy is being violated more anecdotally when you hear from what the meta-employees were saying when they were having their keystrokes monitored. They were talking about feeling super uncomfortable. And also when it was discussed on message boards, the sort of angry face emojis were just pouring out. I think there is a study too, or there is a study under way of some police officers who are often wearing the body-mounted cameras. So all the time they're kind of producing data about their work. And with that being kind of monitored and on the calls that they're making constantly, they don't have those moments to sort of decompress in between jobs, if you like. And they're looking at that kind of really high, intense level of energy and adrenaline happening all the time in the worker's body and how that will affect things. And there was this other study I think, was it the University of Wyoming where I think they looked at people being monitored in this kind of way? And it seemed like they might actually be less likely to comply with all the kind of requirements and actually become sort of more deviant in their behavior. I don't know if that's a protest, but there was another study on that. Yeah, I mean, that's a classic thing, isn't it? When you take someone's autonomy away, they actually end up behaving worse. And what they found there was a paradox really, wasn't it? So it's monitoring paradoxically creates conditions for more deviance. By, I think the thing is diminishing the worker's sense of agency and that that means they're kind of less morally engaged with the work that they're doing. They feel less responsible and that creates a kind of more rebellious worker. So there is a risk that this could massively backfire. If employees just become rebellious or stressed out, but presumably this is just the thin end of the wedge assist technology becomes adopted by more and more companies, where do you see this all going? Well, I think we're already starting to become an issue for trade unions. I think it's going to be a new kind of front line in industrial relations, as well as pay in other kind of conditions. I think that there's going to be also a kind of, I suspect a new dynamic where if you resist monitoring, and you say you don't want to be monitored, does that then start to make you look suspicious to the managers? And so there could be a kind of world in which you have pro surveillance, pro tech, pro monitoring people who are happy to have this sort of almost kind of paternalistic technology hovering over them and people who aren't. And then I think there's the topic that we talked about earlier, which is after a while of monitoring these employees so much using an AI, does the AI then just say, "Do you know what? I could just be doing this myself." You know, I've worked out from these millions of hours of information that I've garnered from these white collar workers that I could just replicate this, and I could actually teach myself how to do this. And, you know, I don't think it will be long until companies start to look at the kind of wealth of data they've got about the way that their employees operate and think, "Well, how can we actually start to use that to our advantage as well?" And then we can all go and lie on the beach and live in the world of. An AI take care of the world for us, right? These dreams. Bring on the future. Bring on the future. Thanks again to Robert Booth. You can read his reporting at the Guardian.com. And that's all from us. This episode was produced by Josh and Chana. The sound design was by Ross Burns. And the executive producer was Ellie Leurie. We'll be back on Tuesday. See you then. This is The Guardian.