How AI gets stress-tested before release

Hello and welcome to the Tech Download, a new CNBC original podcast where we unpack the tech stories that matter most. Each season we dive into one big theme and what it means for your money with insights from the industry's most influential voices. This season we're looking at Google DeepMind, the powerhouse driving the tech giants AI push. Never wondered what would happen if AI got out of control or fell into the wrong hands. And how can governments make sure it's safe for you and I? Our guest for this episode, DeepMind's Senior Director for Responsible Development and Innovation, Dawn Blotswitch and Tom Loup, VP for Frontier AI Global Affairs, addressed these questions. When we think about Responsible AI at Google DeepMind, we're not seeing it as something where it's slowing down progress. Instead it's about us thoughtfully designing it from the very very beginning. Hi everybody, it's Arjen Carple here in London and Steve Kovak in New York. This is the third episode of our three parts into Google DeepMind and I just want to set the scene here because we've we've heard from Demis Osabis the CEO of DeepMind, we've heard from the COO of DeepMind Lila Ibrahim and we've heard about so many topics but this episode we're going to hone in on this idea of responsibility when it comes to developing the assistance. And just to set this scene, there is a growing chorus of voices that are talking about the risks of AI and particularly fears around AI getting out of control as it gets more powerful that the makers of these AI systems won't even be able to control them. The societal impact of AI when it comes to people losing jobs and how it's going to impact all of us. And of course we are placing at this point in time a huge amount of trust in these technology companies, these giant technology companies with vast pockets and there's only there's very few of them developing this technology to get this right when it comes to the safety and the responsibility of developing these systems. And that I think is a really profound thought when you think about it, Steve. Here in the United States, the regulatory system is way different than it is over in the EU. And companies here kind of have the benefit of LACS regulation. We went through this in the Web 2.0 world and the social media world. Nothing happened. We saw all the problems related to election interference, Cambridge Analytica, all sorts of terrible things happening on YouTube and no regulation happened. So when you say we have to put our trust in these companies and the people running them, that's because at least here in the United States, our government has shown a lack of desire to really solve these issues. So these AI companies, they have the benefit of just going at it and the EU, I know you can speak to this a little bit, they're working on stuff in a more serious way. But still, they got some ways to go. And in the meantime, this is evolving faster than regulators can even keep up. Absolutely. And that's the point. The regulators are always playing catch up because this technology is developing so fast. I mean, the tech industry will say, you know, we want the regulation, we want it to be thoughtful and we want it to be adaptable. And the regulators will say, well, yeah, we need regulation, but they also are trying to themselves get to grip with what this technology is. So it's difficult position, but these are the kind of topics we're going to unpack in this episode, both from how do you develop these AI systems responsibly from within? But also how does the regulatory approach fit in here and what kind of regulation could be brought forward? We've got two super interesting guests and I want to kick off with the first one, Dawn Blocks, which executive decisions is the new podcast from CNBC, where I asked powerful leaders about their decisions that changed everything. I'm Steve Sedgwick, here's Mr. Joe Malone, CBE. I started that first business of skincare. That's when I knew that I was in charge of my own life and that's when the entrepreneur really, although I didn't know what the word entrepreneur meant, that's when the entrepreneur really took hold. Now that's executive decisions with me, Steve Sedgwick, get it, we're everywhere listening to this. Dawn, it'd be great to just kick off with, I guess, sort of trying to define what we mean by responsible development. What are the kind of things you're thinking about? We want to build AI to benefit the world and humanity, that's our mission, right? For us, what that means is being really thoughtful about how we're building and how our AI is being used. Our approach always takes into account the AI principles that's guided by that and it centers around these principles of responsible governance, responsible research and responsible impact. When we think about responsible AI at Google DeepMind, we're not seeing it as something where it's slowing down progress or it's about hitting the brakes. Instead, it's about us thoughtfully designing it from the very, very beginning, considering it at all aspects and all parts of the development journey. When I talk about risks too, we think about quite a broad spectrum of risks. You would have seen our Frontier Safety Framework, it thinks about the most severe types of risks, like chemical and biological risks or cyber risks. But we're also thinking about near-term risks as well, so whether that's bias and inequality, because they're really important. We want to think about these things as a holistic set of risks and opportunities. We do see them as very connected, so that's why we don't want to just consider one or the other in isolation, they should be considered as a whole and as something that we are considering in tandem and in parallel. Some would say, Google DeepMind's goal to build AI that benefits humanity, and as part of that, artificial general intelligence, this idea of this extremely powerful AI systems, that there's this tension between being able to achieve AGI and these powerful AI systems and doing so responsibly, that there is always this kind of tension there and actually can the two really coexist. Yeah, I do strongly believe that they coexist and they're not in competition. Going back to the point I made around designing this thoughtfully in from the beginning, if you do that, it's really well considered and it must be done now as opposed to like, we don't want to get to AGI and go, we should now be thinking about these types of risks, but rather, think about it now so that we're getting ahead. A good example of this is the work that we did in AlphaFold, which I realize is not necessarily an AGI example, but with AlphaFold, we embedded in the team for many, many years, working through with them and helping them anticipate some of the opportunities and risks. And that meant that we helped sort of shape it from the very, very sort of like beginning and at the end, we launched it and it was very, very successful and actually saw the benefits that we wanted to see come out of it for four society. And I think that is the process that we want to continue replicating even as we're getting to AGI. The unfortunate thing, I guess, about unintended consequences is that we may not know what those unintended consequences might be. How comfortable do you feel about your level of knowing what the risks are at this point? I think we have a good grip on some of the biggest risks that goes back to the Frontier Safety Framework because these are the big ones that we have considered, but also we've seen considered across the industry. So we know that there is some alignment in terms of the way that people are thinking about that. But we also need to be very mindful and thoughtful about monitoring as well. So looking at exactly how the models are being used and in specific use cases so that we can then say, right, based on this, we didn't see this one coming. And so actually we need to make some changes and some adjustments. And the other big, I think, debate and concern right now is around misinformation and hallucinations with some of these AI systems and granted using them, they've got a lot better, a lot better, certainly from when they were first emerging and getting very popular with users. But again, it's accuracy of these systems is not 100%. So what's the kind of feedback mechanism for correcting and improving upon the misinformation that comes out of the systems that deep minds building? Yeah, I think hallucinations are one of those unintended consequences, I think, of the creativity of the models. And so when we're trying to think about how to address that or where it's where it may be appearing, again, we go back to like whether what we can see the users doing and users will flag things to us if they are seeing issues, but we'll also sort of monitor via the logs. We also have a number of different initiatives that we're putting into place to basically address things like factuality so that we're making sure that we're grounding in the right types of information. Dole, can you help me understand the term red teaming? Red teaming is a really important part of our approach. And so like everybody has slightly different definitions, so I'll give you mine. In our world, what it means is an unstructured sort of exploratory way of testing our models. And we complement that actually with structured evaluations, so day sets of evaluations that we would run as standard on a model. But the red teaming side means we want to have people who either are like real experts in their fields or are wonderful at gel breaking and we want to see them really go in and explore the models and see what they can find. And then they share that with us, we will then make changes if appropriate and then we can also then turn them into structured evaluations. So it's a really critical part of how we operate. So we often hear these sort of stories about people quote unquote tricking AI systems into kind through prompts to give an answer that perhaps that was not intended by the creators of these AI systems. And so is that something similar that what happens in the unstructured side? Yes, so we will have people testing for those jail breaks. So what novel ways can they get the system to say things that we would rather that the system doesn't or it can be in lots of different ways like it can be trying to see whether specific information can be surfaced from a model. So it's all about I think creativity and exploring new ways and every time a new capability comes out. So a model sort of has a new ability. We will then want to test where are the boundaries of that and are there any risks that we see associated with that and that red teaming place a part there. How does the fact that there is such intense competition sort of commercially with deep mind and open AI and other competitors in this field play into what you do because you need to develop quickly but you need to develop safely. Safety and speed they are really necessary parts of a whole we do it together and we know that the competitive pressures are there you know and it drives a lot of progress as well. But it can also mean that people may be tempted to cut corners, but I think that we want to make sure that we are very balanced. We want to keep our focus on the safety and responsibility aspects. We do have and we have created forums so with open AI Microsoft and anthropic we created the the frontier model forum to specifically discuss topics like safety and help us better align on best practices which I think has been really helpful. So when things are moving fast that we have a way that we understand what everybody is trying to do and we have a joint for you on risks which I think is good. Don we spoken about some of the more immediate risks you are talking about some of the things you have to think about longer term what are those for you. Yeah so the two big buckets of things that we're really thinking about particularly when we think about the the agentic era coming into into play is thinking about the frontier safety framework risks but also some of the socio effective risks. So the frontier safety framework includes CBR ends of the chem bio radiological cyber harmful manipulation which we think is very important to consider and then things like loss of control and deceptive alignment. These are all things very broad broad categories of which there will be lots of aspects to explore within that. But the other area socio effective which is about that social and emotional ways that the model model connects with users that to us is going to be a very important area to to understand over the coming year and years I would say. So that includes things like delusions it includes things like relationships and companionship so it's going to contain it's going to contain quite a lot of different topics that we're going to need to explore which will also need to do. In collaboration with a number of different groups including third parties I think civil society academia because we're going to need to agree on some standards there and and there's some really tricky questions that we're going to have to deal with. I suspect you and your team are thinking about how you approach some of the questions around transparency around working with third parties and I do just want to talk about some of those parts as well because there was a letter earlier this year. I'm sure you saw from pause AI and one of the criticisms and allegations they made was that Google DeepMind wasn't living up to the commitments it made in the AI summit in Seoul and the crux of it was that when Google released Gemini 2.5 pro they said there was no safety evaluation that that kind of accompanied it. I know Google and DeepMind have sort of responded to that but there is this kind of broader view I think and it speaks to that there are some groups who believe AI lab should be more transparent with the data going into their models how they're training models and all of these other parts how are you thinking about this idea of transparency. AI is moving so fast we are seeing it as as we've been talking about it in so many different parts of our lives so in knowledge and productivity and creativity is just everywhere and I think in some of these discussions we will find that we're there are going to be some conversations and topics and debates about some topics and I think that is a good thing to have. From a transparency perspective we have we're very thoughtful about releasing information in our tech reports in our model cards to make sure that users whether they're you know customers enterprise customers or developers have visibility of the testing that we have done and where we are confident and where we may have questions but to make sure that they have visibility of the safety testing and the testing that we have done holistically. And we were one of the first companies to release standardized reports in this way and we think that that is something that is going to continue to be important we want people to understand more about the models and we want also for people to rely on on the information that we're providing. Are you sort of constantly debating in how much you should release or can release even just given kind of some of those considerations around well we need to also protect our competitive edge. Yes there will always be that balance with the competitive aspects and that that will go on I suspect but I think that we are always aiming to be as transparent as we can particularly from the safety and responsibility perspective because we want people to understand this is the testing that we have done in frontier safety areas this is the testing that we're doing you know around bias and hallucinations and we we want that sort of transparency and trust. Don't how long have you been at deep mine now nearly eight years seven and a half nearly eight years and you previously were at sales force IDM yes and MPWC is yes what what sort of made you jump into this this world of AI. I've always thought that AI sounded like an amazing space to be in like the possibilities of a really you know the possibilities are huge to be able to support humanity to think about so many different complex problems which we've not been able to do. So this is one of those problems which we've not been able to tackle by ourselves so you know this aspects of sustainability things around medicine and health I think there's so much more we need to understand and so when this role came up to work at deep mind and in the area of ethics and society I was like this is just something I couldn't say no to and very excited I am still here. The bigger question is why did you leave Australia to come to rainy Britain I mean that's my my bigger question it's a question I asked myself a lot to be honest. And don't the other the other part of of that the equation here is you know you have internal processes you have internal teams you have AI principles you have a lot of internal systems in place as you try to develop these AI systems yeah. And the other question is how do you work with third parties and then there'll be a cohort of people say well this is just deep mind grading its own homework effectively so where is that checks and balances. We really value independence as a part of our overall evaluations approach so you know we we have our model development teams doing their their own testing we do testing but also we then have these these third parties. We've been doing this for years is that we wanted to be able to have like an independent view so that we could identify any of the unknown unknowns so we've engaged with a number of different providers very sort of like specialist providers to try and help us understand more about what they're seeing in the models and that's that will be something that we continue to do and we are from a transparency perspective sharing more on in our reports. You know AI labs around the world are sort of coming up with their own set of principles and standards as well but do you think there needs to be something that's more collective or standardized globally I do think that it's it's important to be having these conversations which is why we did have the why we have the frontier model for them because it's exactly that that we want to be able to talk about those standards and how we are calibrating across the the labs on the types of risks that we're looking at how we're actually addressing those. Because we don't want this sort of like completely jagged view of like you know we value something or we were very concerned about something but another lab isn't so those conversations are very very important we really value them and we will continue to do them. Do you think just based on the conversations you've had with your counterparts and some of the other companies or there is a genuine collective sense of responsibility because I think that's that's the other concern from the general public is that you know there's these very powerful companies very rich companies developing these systems that we are all going to be using can we trust them. I think people from across the safety community regardless of lab or you know what company they're working for deeply care about this area that they're very passionate about it and they've been thinking about it for years. I do think that there is a genuine good intention from these groups and a strong desire to work together. I think the safety groups are an invaluable part of how we're developing models and I think they will continue to be. How confident are you now that when a GI is here. - Yeah, that it can be controlled. - We are going to continue to build safety and responsibility into everything that we do, and we've done it, and we will continue to do it. So I think AGI is, in some ways, we don't know exactly what it's going to look like. All we can do really is continue to apply a very scientific approach and have a rigorous approach to how we're thinking about safety and responsibility, and looking at each new capability that comes down the line and addressing them then and there. And I think that will continue to be this sort of like practical way that we can actually address those questions well in advance of getting to AGI. - So your confident sort of that kind of approach will allow you to be in control of those systems? - Yeah. - Dawn, that was such a great insight into what's going on here at Google DeepMiner. I appreciate your time, thank you. - Thank you. (upbeat music) - So it was really cool to get an insight there, I thought, into the way DeepMiner is approaching the development of the AI systems. It sounds like from the very start as they're developing these models, they're thinking of all of these potential risks that could come out of them, and they're stress testing. I asked Dawn there about this idea of red teaming, effectively trying to find loop holes in these systems. But the other tension I thought that was quite interesting here was this tension between tech companies wanting to ship product quickly because it is of course, an incredibly intense competitive environment but also needing to do so safely. - Right, and this reminded me so much of the conversations we're having with about another Google company several years ago. And that's YouTube. So around 2015 to 2017, let's call it. YouTube is going through a lot of problems. And for so long, the messaging of the company was we're an open platform, let anything go, you know, we're not the arbiters of truth. And it took to them so long to realize when you have billions of users watching billions of hours of video a month, you're gonna run into some of these issues and you kind of do have a responsibility to control what's going on out there. If not for just being a good global citizen, advertisers don't like it. And so it was really interesting to see this idea that look, they're thinking about they're doing it, they're red teaming it. It sounds like they're trying to stress test Gemini in a way they weren't doing in the early days of YouTube. So that was refreshing to hear. Are they gonna be problems? Yes, would they admit they're gonna be problems? Yes, we can go on right now and find those issues. But the fact that they're doing it and thinking about it and talking about it right now is quite refreshing. - I said at the start of the episode, we're kind of having to put some sort of trust into these companies to get this right and particularly on the responsibility side. And so we want to know what they're doing and we want to know that they're at least having a good go at trying to make these systems safe from the start. You know, that is one part of the equation. The other part is how do companies like DeepMind and Google more broadly work with the regulators when it comes to thinking about policy around AI? That's a conversation I add with Tom Lou. (upbeat music) Tom, let's just kick off, I guess, with the global regulatory landscape a little bit because it's fair to say regulators around the world are still figuring out AI and what to do with it, how to regulate it, how to legislate it, potentially in the EU there's the AI Act, the US and UK are taking different approaches here as well. So it's quite fragmented. What's the operational challenge for you and for DeepMind as you deal with this kind of fragmentation? You know, I think at the outset, the high level principle that we advocate for around the world is that we balance the need for appropriate safeguards with the need to enable velocity innovation. I mean, the whole point of this is to get this transformative technology out there for societal benefit. Right, and so when you have fragmentation it actually makes it very difficult to scale at speed. And so what we try to do, and I think we've had some success is advocating for more harmonized, globally consistent standards across the board. And some of the things we really emphasize in that kind of approach is to make sure we're thinking about the application layer like regulating the outputs not the inputs. That's really important. You know, not starting from scratch is a lot of existing regulation out there, right? And so you want to be able to tweak, modify, build on that regulation, but you don't need to create, you know, new things out of whole cloth. In an ideal world, what would be your view on the way AI should be regulated in a thoughtful way? And you mentioned kind of, I guess, the word standards. Does that necessarily mean there needs to be legislation or can kind of, I guess, principles that everyone signs up to be enough? The technology and the regulation should be, you know, quite closely calibrated to each other, right? And so you don't want regulation to go out too far ahead and you don't want technology to get too far ahead. It has to be a kind of a constant iterative dialogue among the two and two. I think it's probably gonna need to be a combination of a few of those things. The key thing is to get that balanced right. How difficult do you think the goal of kind of global standards and global regulation is going to be at a time when I guess, countries are kind of almost competing as well in terms of being ahead in certain technologies in particular, things like AI. - Yeah, so these are some really important questions, the global summits. You know, I think they started off as you know, a bludgedly park a couple of years ago, really focused on the area of frontier, AI safety, the France summit last year, really then pivoted to kind of a broader agenda, right? Focusing on AI action and implementation and how we get, again, these benefits scaled to society in a meaningful and transformative way. And then we have India coming up in February. I think the themes that we're hearing that the Indian government really must have focused on are really around global self democratization of AI. And I think those are all really, really important areas that it's great to have these governments around the world really focused on. Now, you ask about how difficult it is to have global coordination and global standards. Look, it's an environment where, in general, you know, I think there's a lot of pressure on international institutions, right? And I think they're over the last couple of decades, I think they're overall strength and cohesiveness has, I think, waned a bit over time. What you're also seeing is a lot of countries put in place measures around digital sovereignty, right? They want to be able to control the governance of their technology. They want to be able to control the direction of the technology. And it's all very understandable, right? It is a very intense competition, both among the commercial players and among governments, right? So I think, as Google beat mine, what we're trying to do is make sure that, you know, in our role as a unique frontier lab at the cutting edge, making sure governments, number one, have the information they need about where the technology is going, how quickly it is innovating, where the area is likely to be disrupted to society, making sure they have that information in a way that they can take into account with their people and their planning. Second is to really work with, you know, institutes like the UK and AI Security Institute, like the US KC, trying to develop a kind of multilateral type of approach towards basic issues that are important with the fact of frontier models working through, for like the frontier model form as well. And then third, I think really leaning into the ability for us to have, you know, coordinated discussions and dialogue really across countries. These summits are a great example. There are many others where we try to have that kind of discussion and dialogue. And hopefully through that lens, being able to get towards a path for more international governance and coordination over time. But of course, it'll be very hard to create a global stand as without China involved in that conversation at a time when, you know, there's clear, I guess, tensions as well as competition between the US and China and other countries as well. And what is, from what you've seen, can China be involved in these global conversations going forward, are they willing to be involved in setting those standards? China has been participating in these global summits. And, you know, it was great to see a number of the Chinese labs. Also, for example, sign up to a set of commitments at Seoul, which is another one of these global convenings that happened last year around publishing safety frameworks. And they have been making some progress on that. But to your point, it is a challenging political environment, right? And it goes way beyond AI. This is around, you know, a broader kind of great power type of dynamics. I do think there are opportunities for discussion, dialogue, in particular, around things like frontier model safety, right? I think, you know, sharing best practices again on how to mitigate for cybersecurity risks, how do you mitigate for things like, you know, harmful manipulation? I think these are things that nations around the world all should be, you know, wanting to address together. The other thing that I think is quite striking about China, and actually about the Asia-Pacific region more generally, is that they are just very optimistic about the technology. You talk to, you know, the public opinion polls. They're very pragmatic about it. They're very optimistic. And there's a lot of excitement about deploying the technology in their societies. And I think you're seeing the Chinese government do. I think a big push towards adoption and diffusion in a very scaled way. I do have some worry that in the US and in some European countries, there is more skepticism about the technology, which I think is going to hinder the ability to deploy in scale in the same way. So I do think there's a responsibility on companies like Google, Google DeepMind, other labs and companies to really demonstrate why it is that this technology is so beneficial for society. I guess it all goes towards this question of, what's the level of understanding like now in government around AI? Does it differ quite substantially depending where you go? Yeah, yeah. As you can imagine, it's pretty jagged, right? There are some governments, I mean I give you one example, I mean Singapore I think is a very, very forward leaning government, very steep in the technology, I mean the conversations that I have with the government, you know, ministries there, it's an extremely high level of sophistication and they are deploying AI very actively and taking a very pro innovation approach. Some other governments, I think for a variety of reasons, maybe as a lot of interest, but also are balancing a lot of competing considerations that cause them to be more nervous or more hesitant about the technology, some of them, I think in general, there's a bit of a risk aversion too that you see with certain types of government officials. And I would say the way that we approach it is, we have to meet people where they are, right? If you're dealing with a country that isn't as forward leaning, isn't as pro innovation in their thinking, our job is to make sure that we do our best to get them into that space, you know, regulators always think about everything that can go wrong, but there's a huge opportunity cost, right? If you don't use this technology to accelerate productivity, efficiency, creativity, right? And I think this is where some of the Asian countries who are facing big demographic crises, I think they're taking a very pragmatic view to those kinds of things and they're really leaning forward and ahead. And so I think we're really trying to get governments around the world really educated about what the technology can achieve, where it's going, and hopefully by doing that infuse them with that kind of opportunity, that kind of opportunity, pro innovation thinking, to get them to really, you know, invest in their technology, invest in the inputs, and hopefully accelerate the ability for that technology to deploy in our country. Don, what's your relationship like, you know, at Google DeepMind with Europe at this point in time, because this is a market that move very quickly on wanting to regulate AI through the AI act. And I know there's rethink happening on this front now in terms of, you know, whether to delay that a little bit, whether to even perhaps tone down some of the GDPR, which is one of the key data regulations passed quite a few years ago as well. But at the same time, you know, the EU has come under fire for perhaps not being innovative enough, but also there's a view, particularly from the US government, that the EU unfairly targets big US tech companies. In the EU, my sense is they are starting to recognize that, you know, too much regulation, too much kind of overlapping, conflicting burdensome provisions that companies have to adhere to is really harmful for competitiveness and innovation, right? So Mario Draghi wrote this very compelling report calling out the fact that, you know, Europe is falling behind because they have this, you know, regulate first kind of mentality. Now, you know, I give them credit, the AI act, you know, ended up in a better place than where it started. I think they are receptive to comments from industry, and we have a very productive dialogue going with Europe and the European Commission, and we're very actively involved in shaping, for example, the implementation of the code of practice and giving inputs. And of course, they have to balance a lot of stakeholders as well, and that that's understandable. But I do hope they are continuing to get the message that European competitiveness from the next generation is at stake. I just want to bring up a story that happened in December with our, for all audience, and this is this EU antitrust investigation into Google, various allegations. Now, I know that's an ongoing thing, but I think it underscores the way, those themes we were talking about, right, in terms of where the regular is stand, where tech companies stand. I know the current EU investigation is very much focused on the way Google is using content online for AI purposes. But I was just wondering from your perspective, Tom, how do you navigate what appears to be very fast-moving potential changes or investigations, et cetera, with regulators around the world? You know, it is a challenging dynamic, right? It is constantly shifting, both in terms of the commercial landscape, the geopolitical landscape, the policy landscape, that's part of what makes, you know, what I do very interesting and every day, I really enjoy doing it. But it is something that we keep a very close pulse on. And I think for us, it's really grounded in those central themes. I was mentioning earlier. I also think, at the end of the day, a lot of this is going to be dependent on the business model that emerges, right? You know, what are the incentives that organizations are going to have and where do those business models lead us to? You know, my firm belief is that, you know, let's take the issue of publishers and, you know, how the ecosystem is going to develop on the web. You know, Google, we are probably the strongest defenders of the open web. We have such a vested interest in having a healthy, thriving web ecosystem, right? And so I think, at the end of the day, you know, a lot of this is going to be solved through business, you know, the normal course of business. We will find a way for there to this to be a win-win equation as we have for, you know, many years across many of our biggest products. And so this is where, again, I worry that there may be unnecessary and maybe imprudent injection by regulators or enforcement officials in a way that is going to distort what I think will be solved by the market over time. Well, I guess content and the training of AI models has been one of the kind of topics that's been thrust into the spotlight quite early on in this AI build out. But what we have seen is some interesting partnerships and revenue sharing agreements between content companies and AI companies. Do you expect a bit more of that going forward? As a strictly legal matter, you know, our position has always been that, you know, the training side of things is governed by, you know, fair use of the US or by the tax and data mining exception in Europe, for example, and, you know, creators have an opportunity and publishers an opportunity to opt out through that process. But I'm thinking more from a longer-term business and practical perspective. I do think there will be, you know, some kind of business solution that's going to be reached, you know, to make this sustainable for long-term. Tom, in 2026 and maybe 2027, what do you expect to be kind of keeping you busy and right get it as busy? So I do think 2026 will bring a new wave of agentic products that are going to, I think, stretch the boundaries, both of the potential opportunities around this, but also some of the risks and challenges that come. So I do think there will be, you know, increased emphasis on things like, you know, privacy and security and new vectors of attack when you come to agentic agents. But also, I think the ability to automate really kind of more end-to-end sophisticated workflows, right? And I think that'll cause some discussion around, well, what does that mean in terms of labor impacts? What does that mean in terms of our ability to augment human creativity and augment productivity as opposed to completely replace it? I also think, you know, personally, I'm really excited about applying AI to scientific discovery. And this is where I think we will have a lot of big breakthroughs that are happening in the near future on this. I think robotics, in particular, is on the verge of a big breakthrough. And then finally, I would also say a lot of this is going to be dependent on the geopolitical dynamics as well. Yeah. That's a side to impact. I think it's going to be really interesting. We saw a lot of job cuts on a quite a large scale, some being attributed to AI, some not. But that debate over the impact is going to be key. And I think that's something policymakers are going to be grappling with big time. Yes. Yes. So for example, the layoffs that have happened this year, you know, I think if you look at them, a lot of them really don't have a lot to do with AI. But AI is kind of an easy way or an easy thing for people to blame. And I do think there is some risk of that, particularly in the US heading into an election year, you know, in the midterms, where AI becomes a bit of a boogie man. And I think again, this is where it's so important for, you know, the companies and the industry really to demonstrate why is this technology so important to invest in? What are the huge societal impacts and benefits that are come with that, right? And what's our, you know, our social license to operate, come to our ability to demonstrate real, meaningful, positive change in people's lives, and I think that's the antidote to that kind of approach. Is that going to be a challenge for you, kind of politicians blaming tech companies? It already is. And you know, I think again, our antidote to that is that we have to really, you know, step up to the plate to show meaningful impact. And again, things like, you know, whether next are, you know, state of the art, whether generation service that is predicts, actually saved lives in predicting. you know, the hurricane cyclone recently, because it has such an accurate prediction in North America. That, those are the kinds of things that I think, which you can really demonstrate that, that real life impact or can make a huge difference. - Great, Tom, thanks so much for your time. Appreciate it. - Thanks so much for having me. Wonderful to have this conversation. (upbeat music) - Arjun, that was a fantastic interview with Tom, and mostly because I'm fascinated by tech and policy and the mix between these two, I think Tom kind of falls into this category where, they say they welcome regulation. We heard that from Sam Altman a couple of years ago. I was in Congress when he was testifying, and the senators were shocked to hear a tech executive saying, "Please regulate us." But it turns out when the regulation comes down, that's when the complaints start. It's like, "Oh no, don't regulate us that way, please." Oh my God, and the lobbying comes in, and the weakening of the laws, and then nothing ever happens, at least that's how it works here in the United States. And then when another country, a major block like the EU does it, the claim is, "Oh, they're just trying to hamper innovation, they're trying to hurt US companies." So again, the trust falls on these leaders of the companies to make it work and work responsibly, and within the bounds of the current law. - Yeah, absolutely. You're right about the European side of the equation. In the European Union, there are a number of pieces of legislation now that regulate these platforms when it comes to data, when it comes to privacy, when it comes to the kind of content on their platform, as well, there are a number of rules. Those are very squarely at this point aimed at kind of big internet platforms. There is legislation here known as the EU AI Act, which is currently kind of working its way through. The Europeans move very quickly on the EU AI Act, and there was, of course, the industry pushback saying, "Well, we really don't know where this technology is going." And I think, to some extent, that's valid, to say, well, we've just come out with some of these products. How do we know where it's going? But also, you can see that, I think it's an example where regulators are seriously concerned about the impact this technology can have if it's misused, or if it's kind of developed in a irresponsible way. And there is this continued tension right now between what the regulators need to do, and them understand is technology versus the companies that are developing in it. And I think that's going to continue. I think, from a regulatory point of view, just trying to understand, where do you even start to regulate the technology? I think the tech companies will say, hey, we want you to kind of regulate the end-use, not necessarily the inputs and kind of how we're developing. And it also, there's technical barriers to company, or to regulators, hey, we need to see the secret source effectively, and we've seen that before. The other part of the equation is, there's always this discussion right now. Well, we need some sort of harmonized global rules. And we couldn't even get harmonized global rules around things like data protection. - Yeah, and look, China has a completely different digital landscape and regulatory regime than we have here in the United States. And effectively, two internets, and you can argue between what we experience in the Western countries and what you might experience in a country like China. And I'll just say one more thing here in the United States. Our president, Donald Trump, gave the AI companies and big tech a huge gift at the end of last year by signing this executive order, basically saying state laws around AI are invalidated, and it has to happen at the federal level. So that's the way our system of government works. We have 50 individual states. They can pass their own laws. And this is basically saying, if California decides to regulate AI, it's invalidated. This is so important. We have to do it at the federal level. - Yeah, Steve, it's a great point. We'll probably wrap it there. It's been so fun launching the tech download with you over these last few episodes. - Yes, and I cannot wait to share what we're gonna work on next, but in the meantime, this is a great way to kick off 2026 just starting with one of the top leaders and top companies and artificial intelligence and getting some real good insights of what we can expect throughout the rest of the year and into the next years to come. - So all you out there, thank you so much for listening and watching this first series of the tech download. We hope you enjoyed it, please reach out. If you have any comments or thoughts on any series or topics you want us to address next. That's it for now. Bye, until next time. (upbeat music) (upbeat music)