Mike Masnick 0:00

Welcome to Ctrl Alt Speech. Ben and I are actually together in person. We are at Trust Con. if you're listening to this, you know it's the world's largest conference dedicated to trust and safety. we have the full live episode recording coming later this week. We're excited for you to hear that. but we just got out of recording a special sponsored spotlight interview where Ben got to speak to Dan Hayes, who is the principal partner for Strategy&, which is a part of the PwC network, and we're going to play that. I think it was an interesting conversation. Ben, what do you... what did you think?

Ben Whitelaw 0:40

Yeah, I thought it was a really, fascinating conversation. Dan's got decades of experience working in the space, and the Outlook report is a really good overview, I think, of where the trust and safety industry is at. one of the core themes running through, as you'll hear, is, is AI, but it also touches on regulatory fragmentation and the changing nature of trust and safety within companies. So there's a lot to get at. Yeah, I hope people enjoy the listen.

Dan Hays 1:11

Hi, I'm Dan Hayes, a partner with PwC in our technology, media, and telecommunications practice. I'm here today with Ben Whitelaw. Ben, thanks for being here. Maybe you wanna introduce yourself real briefly.

Ben Whitelaw 1:23

Yeah. Thanks, Dan. I'm Ben Whitelaw. I'm the founder and editor of Everything in Moderation and the co-host of the Ctrl Alt Speech podcast. Dan, great to have you back on the podcast. Thanks for taking the time.

Dan Hays 1:32

Yeah. Thanks for being here.

Ben Whitelaw 1:33

We found a quiet spot here at Trust Con, the largest conference of trust and safety professionals in the world, here in San Francisco. How has your conference been so far?

Dan Hays 1:41

Trust Con is amazing. Uh, this is, uh, my fourth year here. It's the fifth year of the conference, and it really is the place where the trust and safety universe convenes. Uh, it's amazing. I know we're well over 1,000 people here in San Francisco, uh, and it's a bit dizzying. Uh, quite frankly, there's so many more things that you can see and hear about and learn about here than you could possibly take in. I wish that I could go to every one of the panels and discussions that are here. It's an exciting time, and it's a, it's a really, uh, pivotal time for the trust and safety world as well. I'm sure we'll talk a bit about that.

Ben Whitelaw 2:15

You're here to talk about the PwC's new report, the Trust and Safety Outlook, which is in its second year. Again, we're gonna dive into the themes of that today in this special spotlight episode of the podcast. Before we do, just explain a bit about why you felt that there was a need for a report like this amidst the, the many resources and reports that we see in the industry.

Dan Hays 2:36

The trust and safety world is moving so fast. It's sort of, uh, ubiquitous in the online world, and now it's also becoming even more important as we move into the AI era Uh, where things are even less predictable and less, uh, certain, uh, as if it was even possible. PwC started to, uh, produce the Trust and Safety Outlook a couple of years ago, and really, the, the thesis behind it was we wanna get out ahead of the trends and the changes that are taking place in the trust and safety world. We wanna make sure that we can talk about the changes that leaders need to make in their companies and in their operations, and the things that they should be looking out ahead to to make sure that they are staying up with the latest in the innovations and the challenges in trust and safety. So really, it was a way to aggregate our team's collective knowledge. We work with many of the leading companies in the world who are trying to tackle some really challenging trust and safety problems. And so it's a great way to be able to share that knowledge with the rest of the community.

Ben Whitelaw 3:45

So the methodology of the report is a little bit different to others out there. You, you have a survey of around 2,000 US citizens that you use to kind of ground the insights of the research. Explain why that was kind of important to do.

Dan Hays 3:57

The way that we look at the report, it really has two sets of inputs. One is a practitioner-based set of inputs of what we're seeing out there in the field, working with companies who are looking to innovate and improve their trust and safety operations, and then we also look for a public-based set of inputs that are gathering insights on what are the expectations of- You and me, the common people on the street who are using all of these tools and online platforms, and we wanna hear what they want the online platforms to be producing. We've got a strong belief founded in, in the notion of trust, which is a huge value for PWC. And, and really to be trustworthy, you have to listen. You have to listen to what the public is looking for and what their expectations are. We went out not only to US, uh, uh, stakeholders, but also a sample from the UK as well, and really to hear their concerns and their expectations of the online platforms. It's a great, as you said, sort of mixed approach to being able to help set the course for trust and safety in the year to come.

Ben Whitelaw 5:07

You mentioned AI there, Dan, um, and that's a central theme of the report. You describe it as both the kind of sword that bad actors are using to, to make users' experience online, uh, unsafe, but also the shield that some platforms are using to prevent some of those harms. How does that dynamic kind of play out, do you think, for platforms and, and what challenges does that have for, for them as they u- try to s- use it in both ways?

Dan Hays 5:33

So, you know, it's back to the old saying of fight fire with fire. It's amazing, AI has just ramped up significantly the threats and the challenges that all online platforms have in dealing with bad actors, whether it's, you know, creation of non-consensual, uh, you know, sexual imagery or whether it's the, the, the, uh, constant assaults and fraudulent efforts that are going on. AI has just really, really, uh, increased the volume and the complexity of the challenges that online platforms have. So the fact that they're using AI now increasingly to fight the, those battles, it makes a ton of sense. Now what we're seeing in that space is that AI is becoming, uh, really a, a multi-focus type of approach. First of all, its effectiveness and consistency in many ways can be even better than human content moderators. And so we are seeing that AI tools for content moderation and decisioning can be extremely effective when applied correctly to, you know, what I'll call the basics. They're not eliminating the need for human-in-the-loop type of behavior and for content moderation that has that expertise and that human touch. That's necessary in all of the online platforms. But the tools also are changing very quickly. The tools that are available today were totally different six months ago, and it really is hard for even the most, uh, sort of, uh, experienced practitioner in this space to keep up with all the changes that are happening. And then I'd be remiss if I didn't also mention that AI platforms, AI products themselves are becoming an even greater, uh, source of, of complexity and trust and safety challenges. When we go and we put a new LLM out there, we have to go through a whole evaluation process to understand how is it gonna behave, how is it operating? And we're finding that many of the, the companies now either producing LLMs or even producing more targeted models just simply don't have the experience or the scale yet to go through and do all of those evaluations. And so it, it begs the question of, well, what are the expectations of their users, of their enterprise customers? What do they want those tools to have? What are the guardrails that need to be in place? It's a whole new world out there, and, you know, AI has just really turned up the volume for everyone

Ben Whitelaw 8:02

You talked about consumers' kind of attitudes towards AI there, Dan, and one of the kind of themes that, that come out is the, you know, the feeling of, of general public towards AI agents in particular. This is an emerging area of, of AI and trust and safety, and we're seeing a, a rise of people kind of giving these autonomous agents the ability to perform tasks that they previously would, would have done themselves. I found it really interesting that there was a stat in your report about how 85% of people seemed very au fait with, with the agents and allowing them to do tasks. Was that a surprise to you when you, when you did the research, that that confidence level was so high?

Dan Hays 8:40

It actually is a big surprise. I, I'll tell you, we've participated in a number of studies on this topic now, and my, my first question always is, do those people really even understand what an agent is? Mm. I would argue that most people today have never deployed an agent of their own or tasked an agent to do something. Maybe they've called into a contact center and they've been answered by a, a, you know, a, a, a contact center support agent. In most cases, I think many people don't really even know what agents are about. That aside, the level of comfort with ceding control to an agent was very surprising. Now, what we looked at in the report was, what are the different types of tasks, and more importantly, the levels of risk that are associated with AI agents, and where are people comfortable? There's an emerging school of thought that says the AI agents really will need to be, uh, adapted over time, and in many ways it's us, as the users, who need to adapt. We need to get comfortable over time. So it may start with allowing the agent to read through my emails and give me a summary. That's a relatively low-risk kind of task. But I get a little more concerned when I need to give the agent my credit card and my Social Security number and my frequent flyer number, and I'm gonna tell the agent, "Hey, I need to travel to San Francisco next week. Why don't you just go buy me a ticket?" How much is it going to spend? Is it going to make, book the flights that I would actually want? I don't know those things. So people are rightfully concerned about that. And then when it gets to really sensitive personal information, like making investments on my behalf or making healthcare decisions on my behalf, those things have a long way to go. So there's gonna be a little bit of a learning path for everyone to really understand Okay, I can trust the agent. It operates within boundaries, and maybe there's a point where I'm going to be in the loop and I'm gonna be consulted to, to work on it. But a lot of the products aren't there yet, and so we're, we're just starting to see the early days of how these things are all going to work and how we're going to interact with them.

Ben Whitelaw 10:53

An exciting time, but also a, a fairly kind of nervous one for a lot of users as well. I-

Dan Hays 10:57

it, it can be scary in many ways.

Ben Whitelaw 10:59

Yeah. The report makes the argument, Dan, that the agent should be treated, to your point earlier, with kind of varying levels of permissions and kind of tiers of responsibility. Just wondering what you thought would be the implications of that for kind of a trust and safety team who are potentially seeing a lot of agents be used on their platform or, or their website. How would, how would that change things?

Dan Hays 11:18

You have to look at the agents as, as being multifaceted. So I think for companies that are, that are deploying agents to perform trust and safety tasks, there does need to be a degree of oversight and control and learning, and for everyone to get comfortable with the actions that they take. We've already seen instances in, in platforms today where agents have gone rogue, and they've gone out there and made decisions or taken down accounts or things like that that weren't intended. There's still a lot of learning to be done. Now, on the other side, many trust and safety teams are increasingly being tasked by product organizations with saying, "Hey, we're going to, uh, to put agentic capabilities into our product. How do we provide guardrails to make sure they're doing as intended?" And that's a great opportunity for trust and safety organizations to have a seat at the table and be there alongside product in the design stage and in the testing stage. Um, but they're going to have to learn to be nimble, learn to be very responsive to this because it is moving so fast. It, it really is, uh, you know, depending on what angle you're coming from, uh, as to how you can engage with these things. Trust and safety organizations have the golden opportunity right now. Either they engage in this and, and take ownership, or they risk losing that seat at the table forever.

Ben Whitelaw 12:43

We're gonna come onto the kind of responsibility the trust and safety teams have within their organizations and the role that they play. But just before we do, are we seeing, do you think, a, a, a kind of shift to two versions of the internet? One that is, you know, used by humans, um, you know, in the way that we've always had it since the advent of the internet, and one that is increasingly kind of accessible and, and made possible by agents. And, you know, what are the dynamic shifts that we, we could see when that is the case? I, I feel like, you know, the internet has always been used by humans, and you have a sense that there is a person behind, uh, an account. That's no longer true, is it?

Dan Hays 13:19

Yeah, you could argue that this is a, uh, an evolution that's been happening for some time now. You know, it, it started out years ago as, quote-unquote, machine-to-machine, and then it evolved into Internet of Things. Well, Internet of Things right away implies that humans might not quite be involved in what's going on there. I think agents are the next logical step of it. They're really things, maybe not always physical things, but in some cases they are. They're things that are interacting either with people or with one another. We've certainly seen that happening already over the past year in some fairly, uh, dramatic kind of ways with, you know, uh, agents forming their own religion or agents creating their own language to communicate with one another. These are things that sounded far-fetched a couple of years ago, but now they've really happened. One thing that, that gives me a little solace in this is that the The, the regulatory and lawmaking community seems to be starting to rally around the notion that agents need to be treated in a similar way to humans. Now, I don't mean that you're gonna put an agent in jail. I mean that the person that deploys the agent is going to be responsible for its actions, which means they have to be tested, they have to have proper controls, they have to have been designed appropriately for the tasks that they're doing. That's something that is a great guiding principle for many cases and for many products. We don't have real clarity on that just yet, so that's something we're watching out for in the coming year is how will that regulatory environment, uh, evolve around AI agents as well.

Ben Whitelaw 14:57

That's a great segue, Dan, into, uh, a series of questions I wanted to ask you about. Aside from AI over the last couple of years, regulation has been one of the dominant themes in the trust and safety industry. Platforms having to increasingly do more to comply with kind of government regulation on a whole variety of issues. The report talks about there being a kind of fragmentation across jurisdictions and harms. Talk a bit about what you mean by that and how that has affected some of the platforms and the clients you've worked with.

Dan Hays 15:25

Regulatory complexity is really a plague of the online industry. Every state, every, uh, uh, country, uh, every region, uh, has its own ideas about what, what should be there. And I, I really sympathize with online platforms of having to deal with this sort of fragmented mess of regulations that are out there, let alone having to deal with the societal differences and values that are there. There are countries that simply have different views or societies that simply have different views on what's right and what's wrong. And if you're an online platform, it's increasingly difficult to isolate your audience in different places to have them comply with a different set of rules that are out there. The way we look at this is you have to stay ahead of the curve when it comes to regulation. You have to be sensing what's out there in terms of the emerging regulations that are happening. You have to know, uh, what is the sort of least common denominator for different regulatory, uh, obligations that are, that are present. And you have to, at the end of the day, make some value-based and principle-based, uh, decisions about what you're going to have on your, on your platform. A platform that is enabling kids to play games together online fundamentally needs to have a, a different set of values and principles and policies than a platform that is, you know, maybe, uh, promoting more adult-based content that needs to have a greater, uh, kind of age gating and a greater segmentation of their audience. Those are things that if they can be consistent and if they can lean back on those principles and those values, platforms have a lot easier case in front of the regulators and the lawmakers. When you don't do that, life gets pretty difficult.

Ben Whitelaw 17:18

I mean, naturally, there is this kind of fragmentation that you mentioned, Dan, but we're also seeing a kind of consolidation of regulation around certain topics. You know, we've seen countries kind of corral around a certain view of child safety, uh, the importance of keeping children safe online, as you say. We've seen a number of cases recently, uh, and also, you know, um, kind of jurisdictions show interest in product design as a, as an area that platforms need to be very aware of. How does that kind of change the conversation as well, where we have, you know, a lot of difference in some parts of the, the world around some topics, but- some areas where it's a, there's a lot of kind of convergence as well

Dan Hays 17:56

Convergence in some ways is always helpful, right? And I think even some of the most controversial leaders in the online world would say, "Hey, if the rules are clear, we can operate within them." I do think that some of the product design issues that, that you raise, uh, it's been surprising to see some of those come to the forefront, and I think that they in some ways speak to societal values and societal concerns about the, the growing role that online platforms and AI systems play in all of our lives. Like, you know, one of the ones that we're watching very closely is, uh, this notion of, you know, the, the infinite scroll, right? This has all of a sudden become a real source of concern. And I say all of a sudden because it's now in the public eye. But for years there's been a question about the addictive nature, uh, of, of digital tools and whether or not that's a good thing, a bad thing, how we should moderate it, and how we tr- teach people in their lives, like any sort of potentially, uh, habit-forming product You know, to not eat too much or not gamble too much or not drink too much, right? All of these things have similar criticisms.

Ben Whitelaw 19:13

Mm.

Dan Hays 19:13

So I, I think that for the industry, we really need to think about how to lean on the experiences of others. We don't need to reinvent everything for the online world. What's happened in, in real life is really not materially different than what we're seeing in o- in the online world today.

Ben Whitelaw 19:32

And do you think particularly around product design, do you think there's, uh, are you seeing a lot more interest from the kind of platforms and the clients you work with around establishing a model for that, um, that allows them to, you know, communicate with regulators in a way that shows that they're being responsible and, and adhering to these kind of systemic risks?

Dan Hays 19:51

One of the things we do see a lot of, and it's hard for smaller platforms to do, but for bigger platforms, we see a lot more, uh, access to data and a lot more access to research, um, to be able to share what their understanding of their products and services is. How much time are, are people spending on platform? How do they feel about it? And of course, there are some, some fairly awful cases out there about, about things that have happened in the, in the online world and, you know, we all want to prevent that. As trust and safety professionals, that's core to who we are. But at the same time, there's this balance that's always there to be struck, uh, between personal accountability and the product itself.

Ben Whitelaw 20:36

Mm-hmm.

Dan Hays 20:37

Right? Those are, those are difficult balances to make, and the question is, do we want regulation as a society to sort of overtake our personal accountability? Uh, that's, I think, an age-old question.

Ben Whitelaw 20:48

Conversation that we could, uh, have for many hours, I think. Yeah.

Dan Hays 20:52

Yeah. Ben, I, I know that, I know that you're a relatively new parent. I think you actually had a child during last year's Trust Con. Like, what's your view as a parent? How do you, how do you feel th- view this, you know, personal accountability?

Ben Whitelaw 21:04

I'm very glad that the kind of conversation's being had way before my son becomes a teenager. I'll say that. It's very clear that parents are concerned about how their children are kind of being on-ramped onto the internet, and I can see many ways the, the concern that they have. You know, platforms historically have not always- Um, done due diligence in terms of the features that they've released or demonstrated that they, they have users' best interests at heart, as you say. However, you know, there's a clear sense that people need to take responsibility as well, and that's, that's where I think some of the regulation potentially goes a little bit too far. In the UK, there is a much-discussed social media ban that's coming into force early next year. Whether that happens is, is to be seen. Mm. But the, the people who've been driving that conversation have been parents, not children themselves, who, uh, when asked, have been concerned about the level of, kind of restrictions placed on them. And I think there's a, a tension there that we haven't quite figured out as a society. Who gets to decide what teenagers of different ages get to see on the internet, and how can we make sure that as, as their rights evolve over time and as their capabilities to navigate the internet get better, how does the, the kind of regulation cater for that?

Dan Hays 22:24

Yeah.

Ben Whitelaw 22:24

So I think we, we will see that evolve, for sure.

Dan Hays 22:27

I, I'll tell you, as a, as a parent of three kids who are a bit older than, than yours, my, my youngest is, uh, is, uh, 16 now, I, I've had so many conversations about this with them and, and with my wife, and it's really a difficult conversation as a parent, right? Will you take away this capability from them? Are you okay with them accessing certain types of content? What type of oversight do they need? These are important, you know, parent questions, societal questions for us all to have. It's good that you're thinking about it. No surprise, of course, but- Ah you know, you got a few years hopefully before. But I'll tell you, it w- probably won't be a year before that kid is trying to swipe something.

Ben Whitelaw 23:07

I'm already concerned about how much I'm using my phone in and around him. But it goes back to a kind of central point which I think a lot of trust and safety professionals will be aware of, is that, you know, the level of risk can never be zero on the internet. And I think this, this is the, the point you made about kind of offline historical harms and how we can draw from those. What do you think we need to do as a profession to- educate users that the level of risk can never be zero. Do you think that's something that we, we need to push on a little bit more?

Dan Hays 23:35

The industry as a whole really needs to do more awareness building, and there's a great opportunity for the industry to, uh, to not only educate, but to also demonstrate that it is entering into voluntary standards and voluntary principles based coordination that it hasn't always done. This is an intensely competitive industry. There's a lot of really smart entrepreneurs, there's a lot of really smart engineers who are working across the technology industry, and they all want to win. They all want to have that great market share, be the household name that's out there. But we think that there's definitely an opportunity for the industry to come together in ways that it really hasn't done so much in the past and, and collaborate and even enter into, you know, some voluntary standards. AI and agentic AI is a great example of this. As an industry, there's a wonderful opportunity for us all to come together and say, "These are the principles that we're going to agree that agents are going to operate under," or, "This is how we're going to have large language models operate." Things have been moving so fast that I'm not sure anyone's really taken the time to, to go through that. Mm. And it requires a lot of effort. If you look back into the history of the technology industry, all sorts of great technology innovations, as much as they seem somewhat pedestrian today Many of them were formed by groups of companies coming together, whether it was things like Wi-Fi, where companies came together to figure out, how can we bring together computers that all wanna interconnect and make it easy for users? Or it's USB, where we're all now plugging in our devices and charging them. There's an opportunity for us to really make that simple, but it's gonna require the computer companies and the operating system makers and the app makers to all work together. This is the time where the, where the AI industry in particular can do that and come together with some voluntary standards that are forming up. There's lots of efforts going on out there in the world, but they're very fragmented today. Great opportunity for the industry, and maybe next year at Trust Con we'll be talking about that again.

Ben Whitelaw 25:45

Is there anything you would do to encourage that, Dan? Because you're right. We, we have seen over time platforms work together on, for example, dealing with terrorist content or child sexual abuse material where, you know, these, you know, these harms appear on platforms. The internet is incredibly connected. Harms, you know, proliferate from, from one platform to the next very easily. Is there anything you'd do to kind of encourage that collaboration in, across the industry?

Dan Hays 26:10

One of the things is to think about the value creation opportunity that's there. I'm a firm believer that, you know, in the proverbial, "A rising tide lifts all ships." I think that there's an opportunity here that if we create those standards, it's a- it actually benefits everyone in the industry. So we can't look at it as a cost. We have to look at it as having a real return on investment, and that's something that here this week at Trust Con we're hearing a lot about. It- there's a lot of sessions about, how do you measure and demonstrate the return on trust and safety investments? They can't just be looked at as a cost. They have to be looked at as part of the product, as part of the business case, and as part of the benefits that are, that are happening out there. So th- this is a really important topic for everyone in the trust and safety world.

Ben Whitelaw 26:58

We talked a bit about how trust and safety is evolving, Dan, and, and you kinda touched on that shift again here. The report mentions about how trust and safety as a function should become a more strategic capability that's kind of distributed across various teams within a platform within an organization. Explain why you think that that's important and why, why now is the time to do it.

Dan Hays 27:20

Trust and safety is as strategic or more strategic than just about any function within a, within an organization. Historically, it was a defensive mechanism. It was one where, hey, we need to prevent bad things from happening. We need to manage risk Those things are still there today, but more so we're now seeing trust and safety evolving into a, a place where it's a core part of the product. It is part of the offering. It is part of the value proposition. It's part of a customer expectation. And so, you know, we do view trust and safety as a strategic organization, but it's one that needs to justify and argue for and, and take its seat at the table when it comes to the roadmap, when it comes to the, the product, you know, features and releases, and when it comes to testing, and when it comes to customer feedback as well, right? Trust and safety's gotta be there throughout the entire life cycle. Other organizations have gone through this over time. You know, 20, 30 years ago, you could have said the same thing about legal, where legal was an afterthought for many things, and then people realized, "Well, wait a second, but I'm gonna have contracts. I'm going to have, you know, revenue sharing and all these sort of things. Legal really needs to be a part of how we design things." Trust and safety is right there today.

Ben Whitelaw 28:43

So I think we're seeing some progress in that regard. You know, I've certainly seen larger platforms install kind of chief risk officers and chief safety officers, you know, high up in the leadership chain. And even smaller platforms who maybe don't have that level of hierarchy, they're thinking about safety earlier in the process of building products, I guess, uh, shaped by the recent experience of, of many trust and safety professionals. How do y- you kind of envisage, you know, knowing whether a, a, a platform or an intermediary has, has got to the point where they're thinking about safety? What, what are the kind of... What's the tick box essentially, uh, for platforms- Who, who care about safety

Dan Hays 29:24

Yeah. Well, first I, I wanna acknowledge that increasingly when new organizations and new companies are being formed in this space, they're actually thinking about trust and safety from day one. Um, we're, we're seeing that more and more. We're seeing more trust and safety professionals who are migrating to startups, who are migrating to early stage companies, and taking their experience from, in some cases, larger scaled platforms into smaller operations. That's really great to see. It's a really healthy dynamic across the industry. We're also starting to see, in addition to chief safety officers, chief risk officers, we're also starting to see trust and safety in some cases getting moved into the product organization, and that's something that five years ago would've been almost unfathomable. The fact now that people are viewing it as part of product and, and that policy is part of product i- is actually really exciting for the industry. Is that gonna be the ultimate form of organization? I don't know about that, but it's great to see that acknowledgement. And then we're also seeing, you know, fewer questions happening about, "Hey, why are we spending this on trust and safety? Why are we doing this?" We are seeing the logical questions about how can we be more effective with our spend? How can we be efficient? And that's where a lot of the things that we were talking about earlier around automation and the use of AI and trust and safety are, are really starting to come to the forefront. It's important for trust and safety organizations not to rest on their laurels. They need to be thinking about those things and thinking about it as part of the business. I

Ben Whitelaw 31:02

mean, one of the side effects of that efficiency drive of using AI has been that some platforms have, you know, relinquished human trust and safety experts. You know, they've, they've got rid of, um, people and there has been layoffs across the industry. Professional development is a kinda key strand within Trust Con. There's a lot of discussion about what the future of a trust and safety professional looks like in a world that's increasingly kind of, you know, automated where agents maybe are playing a larger role in the process of reviewing content and assessing appeals. How do you see that, that professional dev- development journey changing, and where do you think kind of the trust and safety professionals that exist nowadays are going to end up?

Dan Hays 31:43

I think professional development is a hot topic for everyone today, not just even trust and safety, right? There's widespread disagreements and, and debates about the effect that AI will have on the economy and jobs in general. Trust and safety is not immune to that. I think what is changing is the paths for people. Many people in trust and safety started out in, with the very basics. They were doing content moderation. They were making decisions and implementing the policies and the processes that were there. With AI taking a greater role in the fundamentals- It's probably gonna be more important for people in the trust and safety community to really think about and understand this is how things work, this is how the AI tools operate. But they may not have that foundational first-person experience in how all of it works. So increasingly, trust and safety professionals need to know the engineering skills about using AI, they need to know the policy skills about how policies can be set out, and how to make them most effective, including for how they're gonna be used to train the AI systems that are being used. So it is reshaping that dynamic of pr- of career development for a lot of trust and safety professionals. I think the trust and safety community has largely risen to the challenge. We've seen just tremendous growth, and a lot of people embracing the tools. Where it is causing challenges, and we talk about this in, in PwC's Trust and Safety Outlook, it is causing a shift from what I'll call more of the business process outsourcing-led model, where you had thousands upon thousands of humans, uh, doing content moderation, to now, a, an AI agent-led model where that same volume of humans probably won't be there, at least for the most basic of tasks.

Ben Whitelaw 33:36

Mm. Yeah, that certainly seems to be the way things are going. That question of kind of- Technology vendors and, you know, B- BPOs that sit around platforms that have enabled a lot of the content moderation at scale over the last two decades. That certainly is shifting, and I wondered kind of what you thought that might look like in five or 10 years' time. These companies play a very important role in shaping policies and, and enabling teams to kind of work in the ways that, that we've talked about. But some platforms have decided to kind of, you know, move away from that vendor model. We've seen a lot of startups kind of, you know, be acquired or, or, or merge or certainly kind of partner a lot more. What do you imagine that ecosystem to look like?

Dan Hays 34:18

We see two vectors happening within that ecosystem of, of vendors. The vendors who are incumbents in that space and who are playing in that space are, are being really pressed to develop their own often AI-led capabilities that will supplant many of the human capacities that are, that are there today. So they're having to reinvent their own business models in order to really make this happen. Uh, you know, in, in addition, there's a, there's a general need to shift the human skill set towards more expertise, right? It used to be the really basics, right? I look at an image. Is it violating? Yes. No. I'm done. I'm onto the next one. But now it's gonna be much more of a nuanced expert-led type of model. Whereas you used to have someone who was perhaps working for minimum wage in the space, and maybe they, the average person only stayed for a year and moved on to another career, there's an opportunity now to retool this as an expert-based type of model and for content moderators to be playing the role of the human-in-the-loop expert that are dealing with the corner cases, that are dealing with the complex cases that, that AI really isn't going to get to anytime soon.

Ben Whitelaw 35:38

And I wonder if it will change the way that people and the general public think about content moderation and/or think about frontline moderators at least as that model shifts. You know, we've seen AI companies hire kind of PhDs, uh, and a- experts, as you say, to train models and evaluate data sets. And there's a, there's an interesting kind of dynamic there of, of the old model still existing, but a new model emerging.

Dan Hays 36:02

We're absolutely seeing that today. I mean, some of the most interesting things going on in this space are the, the level of expertise needed to train the AI models and to do evals, uh, is really deep these days. So we're seeing clients of ours who are out there retaining, for example, medical professionals or financial professionals who need really nuanced, really deep levels of expertise in order to help with model training, in order to help with some of these more complex evals that have to take place. It, it's a brave new world when it comes to, you know, trust and safety.

Ben Whitelaw 36:39

So I wanna wrap up today's conversation, Dan, by asking you If you are an attendee of Trust Con, perhaps listening to this on the way home, uh, out of San Francisco, what would the takeaway be for them from the Trust and Safety Outlook Report? What do you think that they should be thinking about? What should be top of mind for them and their teams as they head back into their kind of working week next week?

Dan Hays 37:01

I think there's gonna be a lot of positive energy coming out of Trust Con here. And, you know, the sessions are just unbelievable in terms of their depth and their breadth. Uh, you can learn about just about anything in the trust and safety world here at Trust Con. For me, the real takeaway is a- as we all go back to our day jobs, as we all go back to our companies, we have to be thinking about the ways that we're gonna add value in the year to come, and how to demonstrate that value and communicate that value to all of our stakeholders, whether it's our boss, our other functions in the company, whether it's to the, the users or the customers of the company. It, it's important for everyone to be able to depict that value and to not stand still. We've got a real opportunity to move trust and safety ahead in a meaningful way, that the technology has made it easier than it's ever been, and at the same time, more complicated. So we've got a lot of, a lot of opportunity, but there's a lot of work to be done.

Ben Whitelaw 38:03

This has been a great conversation, Dan. Thanks so much for taking the time to explain some of the key findings from the report. Um, I'm sure many people from Trust Con will be itching to get their hands on it. Where can people kind of read the report and get more from it? Where can they find out?

Dan Hays 38:17

PWC's Trust and Safety Outlook 2026. You search it up, it'll pop right up on pwc.com. Uh, but we're really looking forward to, uh, to sharing it, and we'd love feedback on it as well. Uh, I'd encourage anybody who takes a look to please send me their thoughts. Uh, I'm easy to find online and all sorts of social platforms. And, uh, yeah, look forward to it. And Ben, I really appreciate the opportunity to join you on the podcast today.

Ben Whitelaw 38:41

Well, thanks so much for joining us today. Um, it's been a great conversation. Uh, it's been a pleasure talking to you, Dan. Take care, everyone.

Announcer 38:49

Thanks for listening to Ctrl-Alt-Speech. Subscribe now to get our weekly episodes as soon as they're released. If your company or organization is interested in sponsoring the podcast, contact us by visiting ctrlaltspeech.com. That's CT RL alt speech.com.