Who’s the sheriff in AI’s wild west? (Equity Live at TechCrunch Disrupt 2026)

The Equity crew was live at TechCrunch Disrupt 2026! Kirsten Korosec, Anthony Ha, and Sean O’Kane took over the Builders Stage Tuesday morning to tackle a growing challenge for the AI industry: what happens when AI agents go off-script, and who’s responsible for keeping them in check? Recent incidents, from Anthropic pulling internal evaluations off the live internet after agents misbehaved to Microsoft CEO Satya Nadella calling for an “emergency brake” on AI models, have brought agent security into focus. The Equity team unpacked whether the problem calls for better engineering, new regulation, or both, and why startups are raising millions to secure AI agents. Listen to the full episode to hear more about: The Anthropic agent mishaps, including a false police report that went nowhere after landing in a spam folder Nadella’s call for an AI emergency brake, and why the hosts wonder what took so long The funding race to secure AI agents, from AIR’s $50 million raise to Reco’s $55 million and Oak’s $60 million for agent identity, plus Sequoia-backed Cymphony Why independent verification and certification could be key to building trust, and whether Big Tech will buy its way out of the problem Subscribe to Equity on YouTube, Apple Podcasts, Overcast, Spotify, and all the casts. You can also follow Equity on X and Threads. Chapters: 00:00 Intro and what we’re excited for at Disrupt 01:34 Durability, security, and Anthropic’s rogue agents 04:41 Nadella’s “emergency brake” post 06:33 NVIDIA’s OpenShell and the engineering-vs-regulation debate 10:04 The agent security startup boom: AIR, Reco, Cymphony 13:38 When agents go rogue in real life 15:55 Certifications and standards 19:33 HiddenLayer and the jump to production 20:40 Will Big Tech just acquire these startups? 22:52 Consumer trust and agent adoption 25:05 Predictions: acquisitions vs. independent verifiers 27:26 What’s coming up at Disrupt 29:02 Outro Learn more about your ad choices. Visit megaphone.fm/adchoices

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Speaker 00

  This episode of Equity is brought to you by Schneider Electric. Schneider Electric, advancing energy tech.

• 0:17Kirsten Korosec?

  Hello and welcome to Equity, TechCrunch's flagship podcast about the business of startups. I'm Kirsten Korsak and I'm joined, as always, by our weekend editor, Anthony Ha, and senior reporter, special project, Sean O'Kane. We're here alive in person, exciting.

• 0:34Speaker 00

  Super exciting.

• 0:35Kirsten Korosec?

  Yeah, yeah. What are you looking forward to?

• 0:37Speaker 00

  I'm definitely looking forward to the Flock interview tomorrow. I know we have R.J. Skarringe onstage today, right?

• 0:44Kirsten Korosec?

  Yeah, in a couple hours.

• 0:45Speaker 00

  Very, very soon, yeah. So I would say those are two big ones, and then also the startup battlefield. Startup battlefield is a big one, and of course, like, it's a bunch of founders who are very excited to be here. Companies you probably haven't heard of before, they're going to be going toe to toe on the Disrupt stage.

• 0:58Kirsten Korosec?

  Yeah, what about you, Sean?

• 0:59Speaker 02

  We've got a big space gun. I feel like we're varying the lead. There's a big space gun over near the main stage. Now, I mean, there's a lot of space stuff, which is exciting for me, a lot of satellite stuff, which is really cool. Obviously, mobility. It was such an interesting crop of startups to claw through this year as we got ready for the show. I think the theme, unsurprisingly, is AI, but more specifically, building durable businesses with AI, durable startups, and trying to find a path forward. And I'm excited to have a lot of those conversations this week. I think as we take a look at what's happening in the news, though, we have a lot of questions that still need to be answered about how that technology gets used, right?

• 1:39Kirsten Korosec?

  Yeah, I mean, it brings up like the idea of like, what is durability? And to me, a key piece of durability is security. And we have a lot of examples, even just a few days ago, of AI agents kind of going rogue. I mean, there is the Anthropic situation that just happened. There's so many situations, though, right?

• 1:58Speaker 00

  Right, I mean, it seems like the last few months have just been this cascading series of headlines where it's just like, oh, OpenAI admits that like this AI agent, you know, did this thing that it didn't want. And then Anthropic has an announcement, and then OpenAI has another announcement. And so in this case, there are a couple of big pieces. One was that they had reported that there was a, they filed a false police report. And that was just, I guess, one of a whole cluster of different rogue AI agent activities. And so what Anthropic was trying to say was, well, if I guess, first of all, oops. And then, secondly, so moving forward.

• 2:30Speaker 02

  That's a technical term, right?

• 2:32Speaker 00

  Yeah, that's what we're here for. That's what you need the journalists for.

• 2:34Kirsten Korosec?

  By the way, not to interject, but like, that police report thing, the thing that like caught my eye was the fact that nothing came of it because it was caught in a spam folder, which was like a really lucky.

• 2:44Speaker 02

  It always comes back to email.

• 2:46Speaker 00

  You know, then the flip side of that as well, if there's more and more things like that, then maybe more and more legit things start to go to the spam folder. So that still can do potential harm. So at least what Anthropic is doing for the near term is, we're just going to, basically, all, when we're testing our agents, we're not going to connect those computers to the internet, essentially.

• 3:04Speaker 02

  Which seems challenging, right? I mean, these things need to be able to access the internet to be able to do a lot of the tasks, especially as we talk again and again and again about agentic AI and being able to do things for you. They need to be able to train on that stuff. And I think there is a bit of a question right now of, okay, it can be good from a safety perspective to cut off that access, but what is that going to mean for the models that end up getting released? How are they going to approach training it? I do want to say really quickly, the police report was submitted in Philadelphia. I think this was revenge for Pitchbot. This was AI coming back and saying, you murdered our robot friend a decade ago, and now we're going to mess with your police department.

• 3:45Kirsten Korosec?

  Interesting conspiracy theories here on this stage at Disrupt.

• 3:48Speaker 02

  It's the first of many.

• 3:50Kirsten Korosec?

  Anthropic isn't the only company that is dealing with this. I mean, we've seen a series of these, and it really gets to me thinking about what needs to be happening at the front end instead of this sort of reactive period I feel like we're in. And I don't know if you two have that same sense, but just since the summer, really since the hugging face incident with OpenAI, it just seems like a snowball effect is occurring where it's just, we're still catching up. We're still catching up to how to get ahead of this. And then there's, of course, this debate of, is this an engineering problem? Or is this something else, a little bit more, you know, doomer-esque, that's playing out in the industry right now?

• 4:31Speaker 00

  Right, so just as another example of the discourse around this, over the weekend, Microsoft Satya Nadella, the CEO, he put out this kind of long post where he talked about the idea, again, these sort of, I guess, we're kind of in this season of like, AI executives putting out these long posts, which is like, these are my proposals for how we can fix this.

• 4:49Kirsten Korosec?

  It feels like 2010, like the blog, CEO blogs are back now.

• 4:52Speaker 00

  Right, and I think like, we should talk about the substance of what he said, but also zooming out slightly. I think part of the reason why they're putting out all these proposals is because they're like, if we don't have our own plan, then that plan is going to be imposed on us by, you know, state governments, probably not the federal government, at least, we'll see. But I think that, and very notably, his plan doesn't say anything about regulation. It's about, here are some technical fixes that might give us more control over these agents in the future.

• 5:18Speaker 02

  Am I crazy? Like, why are we talking about this now and not a year ago? Like, these models were definitely worse a year ago, but they were still able to do some of the basic things that we're talking about here, like accessing the internet, you know, doing some of this basic work, and getting in position where they could veer in a bad direction. They have that freedom, even in the public versions that were released to consumers. So I feel, I feel crazy that we're talking about this, like, you know, it's just like one of those situations where you've got the ultimate CEO post, you're right, it's like a very CEO blog from like the early 2010s of like, where have you been? You know, was Satya too busy trying to make Google dance? That like he forgot to check the models that they were training?

• 5:58Kirsten Korosec?

  Yeah, I mean, the quote here is, "We must assume a model is compromised and contain it from the start. Think of it like an emergency brake." And, you know, as a transportation editor, I like to think of emergency brakes and the components of that being put in place on the vehicle prior to it being sold, not after.

• 6:15Speaker 02

  It's aftermarket.

• 6:16Kirsten Korosec?

  Right, right. Well, I will say that there are opportunities then, as a result of this sort of reactive period. And Nvidia, I wrote a story just a few weeks ago about, you know, their system that is very largely much open source based on reining in these sort of rogue AI agents. And Jensen Huang, the founder and CEO of Nvidia, is very much falls in the, this is an engineering problem side of the debate. And they opened OpenShell, which is this idea of creating these layers that will provide these security layers, the emergency brake, if you will, and will allow for the kind of testing that you want to do, and not divorce it from the live internet, which I think is an important piece.

• 7:01Kirsten Korosec?

  But to me, what's really interesting, because we're here at Disrupt with so many startups, Nvidia, the giant, is not the only company trying to address this problem. There's a ton of other startups that are, you know, really tackling this emergency brake piece of it or a security piece of it.

• 7:17Speaker 02

  I mean, if it's an engineering problem, I think we need to think about whether or not those people are going to be held responsible if things get really out of control.

• 7:25Kirsten Korosec?

  Yeah.

• 7:26Speaker 02

  And I think that's one of the reasons that this conversation gets so complicated, because the closer you get to treating it that way, the more likely you are to be regulated or held accountable for the actions of the software that you're creating. Again, certainly not at the federal level right now, but I do wonder at what point, you know, what state attorney general is going to do that. Startups don't necessarily have to worry about that as much right now, especially the ones that we have here this week. But when you're looking for guidance as to where this stuff is going, you're looking to those bigger companies and looking to see whether or not they're doing it. One thing I would love to know is, are they going out and buying insurance against that? You know, like, are they hedging in some way for an eventual future where they're going to have to deal with that?

• 8:05Kirsten Korosec?

  Well, I mean, I'll take the free market position on this and, you know, as a capitalist, and say, sure, there might be a regulatory piece to this, but what about just, you know, companies that are not addressing this problem will fail? I mean, is that a fair enough point? Or is it, you know, some of these companies are far too big that we need to have a regulatory stance in controlling sort of the security component?

• 8:27Speaker 02

  I mean, it's just one of those situations where it would be nice to have an actively engaged regulatory scheme at the federal level so that we can figure out some of the answers to those questions.

• 8:36Speaker 00

  Yeah, and I mean, I do think it is important to remember that all the CEOs, like, sort of making this pitch for either no regulation or extremely limited regulation, like, that is in their interest. And they want to frame this as an engineering problem as much as possible. And I think it is that, but also, we all, everyone here also has a stake in the outcome. And, you know, the free market is not always the best way to sort of sort all of that out. That said, and then, well, I think another question then is, to what extent is this a problem then that a startup outside one of the frontier companies can solve versus like, they have to make it safer themselves because they're the ones building these models? Like, how much can you make it safer when you're just reacting to changes in these models? That said, I mean, our colleague, Ram, I think, looked at Crunchbase and PitchBook,

• 9:21Speaker 00

  and he saw like, two dozen companies selling some form of AI agent security. I mean, that's almost certainly an undercount. That's just like what he was able to find on a quick search. So clearly, there are companies getting funded in this space, and that makes sense.

• 9:32Kirsten Korosec?

  There are a ton of them, and there are a ton that we have covered here, since we do love our startups. And a few of them, like Air, they raised $50 million to vet the tools that agents use. Of course, then there's like Rico, which raised 55 million. And as an AI agent, security startups sort of effort. You know, the interesting thing to me is the stakes are higher for these smaller startups. Right? And OpenAI can afford a little bit to be reactive because they're well capitalized. But these companies really, not only is there an opportunity, but they have to be sure that their product is sort of working and they can't react to things.

• 10:14Kirsten Korosec?

  So, you know, credit to all the startup founders out there that are really trying to get ahead of this issue.

• 10:21Speaker 02

  Yeah, I mean, again, this is another interesting sort of hedge that we're seeing where Sequoia is backing one of these companies, Symphony. And you're thinking about protecting yourself on like letting startups get creative, find ways to help solve this problem while also backing the biggest companies that are, I don't want to say creating the problem, but where these questions are being generated from. And it's, you know, that's good on the investor's side. And there's so much capital to play with now, especially among some of those bigger investors and these biggest firms, that they can play all of these pieces. You know, it's really like, you know, it's like protecting yourself on the downside risk of like a public market trade. Like, it's kind of how I think about it. And I think that's really interesting.

• 11:03Speaker 02

  And I think that's a really exciting place for startup founders to be because you can say, I don't want to go after this big problem that has big thorny questions, but I think I have a really clever idea how to hedge against it in this way.

• 11:17Kirsten Korosec?

  Yeah. I wonder how many slices that you can make, though, within this? Because we have a nice little list here that our producer compiled for us based on our coverage. And the list is endless. And I'm like, how many more, like, how thinly can we slice it? How many startups can exist just addressing this issue alone of being less reactive and with security first?

• 11:40Speaker 00

  Right. It's probably not a winner take all type situation, arguably compared to some other areas in AI. But this is where you can imagine that different companies are going to want to work with security vendors. I mean, because a lot of these, where the security comes in is right, they're being bought, the customer is the company developing this model or building agents. And then they're like, oh, yeah, like, I'd prefer if this agent doesn't go wildly out of control and commit crimes. And so let's buy this software to make sure that doesn't happen. And so, yeah, I don't know that like, there's necessarily going to be 20 of these companies, you know, a few years from now, but I could imagine there being three or five that are successful. So it doesn't, yeah. And then how you slice it, I can imagine you probably getting into stuff that, you know, somebody who maybe knows a bit more about security. But I'm also curious about how much it's going to be new startups and how much it's

• 12:25Speaker 00

  also, you know, on this list, we have CrowdStrike, you know, other companies that have been around for a while. But if you're a security company now, obviously agents are going to have to be a huge part of what you're thinking about.

• 12:35Speaker 02

  I know it's not necessarily a problem for most of the consumer models. And we're talking more about the ones that are being trained. But have you guys had, have you used anything that has felt like it has gone a little rogue yet?

• 12:47Kirsten Korosec?

  I haven't had some of the like high profile incidents of like something booking something for me. And, and the famous one, of course, is the gym reservation. And I haven't had anything like that. But I'm also naturally suspicious of getting up. So, so I tend to limit sort of access. But you've been playing around with Muse a little bit.

• 13:08Speaker 02

  Yeah. It hasn't really gone rogue, I would say. It's definitely, I think the strangest experience using Muse is when it is just like sort of very plainly showing you how it's clicking around. I mean, we're back to like, sort of what we saw happening in the back of, back end of like the Rabbit R1 and like lots of these other sort of automated software systems that were being pitched as agentic like two years ago, but weren't really quite there. And so it's funny to see it like break that illusion where it's like literally dragging a mouse around a screen in the app on your phone. The only situation where I've come across something that has made me feel a little worried is actually in the course of doing our job where like, I wasn't even meaning to use an AI agent of any kind, but I was calling a police department to try and find more information about a police report.

• 13:53Speaker 02

  And whatever front end voice agent system they were using took me asking for information about police reports at a certain location as me reporting a location, the problem at that location. And it ended with the voice agent telling me that it was going to send somebody to that, that location. So I very quickly had to get back on the phone and find a human and say, please stand down. I do not want to be involved in this. And that felt a little freaky.

• 14:20Kirsten Korosec?

  Well, so that I'm sure there's a startup out there, you know, working on that sort of identity issue, right? I mean, I think Oak is backed in like 60 million in funding. They just came out of stealth to fix the identity mess that AI agents are making. So you see a lot of these opportunities. It does feel right now, though, as I said, you know, just a few minutes ago, this sort of reactive moment where we're not quite ahead yet. And I'm sure that there are maybe some startups that are here this week that are, you know, tackling that issue.

• 14:54Speaker 00

  Yeah. I mean, I think what will also be interesting to see is whether there are use cases that start to emerge or people starting to talk more publicly about approaches that you said that are more proactive, right? Because right now, it's mostly just like, oh, we screwed up. Like, it's just this, you know, one after another. But if companies can actually say, hey, you know, we are much more secure and we've done this testing. I mean, I imagine that probably another thing that will have to emerge for those stories to start to happen is some kind of standardization around kind of like certifications and things like that and ways to be able to say, hey, a third party has verified that my agents are not out there doing bad things.

• 15:35Kirsten Korosec?

  Yeah. In the nerdy transportation world, there are like these third party validators and certification that you can have that meet like certain technical and security quality measures. And I kind of wonder if that will happen in this realm.

• 15:52Speaker 00

  Every company today is looking for an edge. Better insights, faster decisions, greater resilience. But as AI drives new growth, it also creates new demands on energy and infrastructure. Schneider Electric is connecting energy, automation, and digital technologies to help organizations make smarter decisions and operate more efficiently. Because the future won't be shaped by AI alone, it will be shaped by energy and industrial intelligence. Schneider Electric, advancing energy tech.

• 16:18Speaker 02

  I mean, you always learn the risk of, what's the SKCD comic of like, we have too many standards. We need to come up with a way to simplify this. We now have one more standard than we already had. Like, there's always the risk of that. I do think there's an element of, maybe I'm reading between the lines here, but if you read Satya's post that we talked about, I think he's like aiming in that direction of some of those ideas. But again, there's this, there's this like trust fall gap of the companies are enjoying the fact that they don't really have to do some of that work right now, which is like, I personally think is why we're seeing so much reactive versus proactive. And so they're, they're just trying to luxuriate in that as long as they can. And who can blame them, right? Like, if you're, if you're not being regulated, you take advantage of that as much as you can.

• 17:03Speaker 02

  But he's aiming in that direction. And so it really turns into a situation of who will stand up and start to come up with some of those standardized things that, you know, he makes some good points in that. Like, he makes a point in that post about like, we need to be careful not to have, you know, we can set up these monitoring systems when we're testing, but we don't want to turn it into like a black box nested in a black box. Like, he's aware that this could spiral, you know, just putting raw effort into more monitoring could turn into a situation where it actually gets worse.

• 17:31Kirsten Korosec?

  Yeah.

• 17:32Speaker 02

  And so how do you balance that out? Then it just becomes a question of who actually takes that step.

• 17:36Kirsten Korosec?

  Yeah. I mean, right now, I mean, this terrible analogy, but in the era of the Wild West, which was a very short period of time and incredibly chaotic, and there was a lot of mayhem, and then like the pendulum swings to like, you don't want the pendulum to swing so much to the other extreme that it like wipes out innovation, right?

• 17:52Speaker 02

  My God, now I'm thinking about an AI Pinkerton agency.

• 17:56Kirsten Korosec?

  Yeah, don't go there, sorry.

• 17:58Speaker 02

  That's all I'm going to think about the rest of this episode.

• 18:00Speaker 00

  Yeah. And I think, you know, Sean, one thing you were saying earlier is that obviously this is a situation, this kind of Wild West era is a situation that these AI companies have taken advantage of in a lot of ways. You also use the word luxuriated. I think probably they are no longer luxuriating in it quite as much.

• 18:18Speaker 02

  Maybe.

• 18:19Speaker 00

  I think they're a little concerned at this point, which is why you're seeing the flurry of like, you know, posts and meetings. Because I think there's a little bit of a tone of, is this enough? Is this enough? Is this enough?

• 18:32Kirsten Korosec?

  Right.

• 18:32Speaker 00

  And, and so, you know, I don't know that it's necessarily the number one thing on their minds, but I think it's, it's, it's up there.

• 18:38Speaker 02

  It's like my six year old who's starting to really enjoy testing whether he could swear or not.

• 18:43Speaker 00

  Yeah.

• 18:44Speaker 02

  You know, and it's just like, he'll let it out and you can kind of see his eye trained over to see if he's getting a reaction.

• 18:49Speaker 00

  Yeah. And I mean, I think on that question of kind of what does it look like to move from reactive to proactive? And also this question of like, why weren't they more serious about this before? Because that was also my thought, like reading Nadella's post where there are things where, again, like he talks about the emergency brake and like, you know, there should, there should always be somebody who knows how to like turn the thing off if it's starting to go bad. And definitely my first response is like, yeah, obviously. Like, why, how could that not be something you're thinking about before? But speaking about that transition, there was this great quote that the founder of this company, Hidden Layer, shared with us, which is this idea that when agents reach production, the scale of their cost and risk goes from theoretical to full scale really quickly. And so like over 50 of his customers of AI agents in production touching critical systems and sensitive assets.

• 19:34Speaker 00

  And so I think that speaks to the just like how we are like at this really, really critical transition point where we're going from, hey, AI is this thing I'm trying out to AI is this thing that's core to my business and I could really screw up a company's business if it doesn't do what it's supposed to do.

• 19:51Kirsten Korosec?

  Well, it goes back to sort of our theme, which is durability. I think this is the startups that can solve that and sort of, again, be the answer to these companies that are reacting. Not only does it make the larger company that they might be serving as an enterprise customer durable, but it also helps them either raise funding or exist as a company and become more durable. To me, what's really going to be interesting is not just how many little slices or little fragments of this sort of security question, different startups, but then how many of these startups end up getting acquired? Because I don't know if they're all going to fold or fail or whatever. I actually see a lot of opportunity for the Microsofts of the world to go,

• 20:36Kirsten Korosec?

  oh, that's an interesting startup. Oh, that's an interesting startup. Let's just fold this in instead of having them as a customer.

• 20:43Speaker 02

  Well, and that's where we come back to, what do we call it? The morally binding agreement that we saw at the White House? Like, are we in a situation where if a bunch of startups come up with clever ideas that can help the bigger companies rein in this problem, and they get acquired by those companies, are those companies going to be incentivized in any way to share that approach or that knowledge or that tool or whatever it is? Because you look at that framework agreement, whatever you want to call it, it does not seem like it is really putting pressure on any of these companies to do something like that. They all say that they want to be able to sort of match each other on these capabilities. But, you know, when push comes to shove, we're already seeing

• 21:26Speaker 02

  a differentiation, let's say, between how they're all approaching the speed that they're operating at. Like Anthropic is talking about pacing the frontier. OpenAI is kind of talking about taking a more cautious approach. But then we have Meta, you know, releasing Muse and then releasing its sort of like frontier versions of Muse this week and talking about, no, I'm going to take the foot off the gas. If you have one company doing that, let alone a split of a few, that balance is never going to get struck. And so like, it's just hard to imagine a world in which they're all going to sort of work together.

• 21:55Kirsten Korosec?

  Well, there's nothing like one big company hitting the accelerator, so to speak. All the other companies are going to then want to respond to that. They're not going to kick back and relax and go, yeah, you got this.

• 22:05Speaker 00

  Yeah. I mean, I think in terms of durability, I feel like security in general is an area where, you know, risk and opportunity are sort of intertwined and the risk is kind of forever, right? Because in this case, the agents, the models, those are going to keep advancing. So you're going to have to continue developing new technology to control that, to respond to that. And again, this is where the engineering challenge meets kind of the bigger world is, is the attention that this is getting going to be durable? Or is this something where they will reach a point where they feel like, okay, like people are freaking out about something else, so we don't have to worry about this as much? Which is possible. I do think that certainly over the next year, and this is beyond just like enterprise, but I think it affects consumer AI adoption as well.

• 22:48Speaker 00

  Because if all you're reading day after day is like, this AI agent has gone and like submitted like false police reports or what have you. We've talked before about how like, you know, sometimes this kind of doomerism can be a kind of marketing because it is like, wow, like this is-

• 23:02Kirsten Korosec?

  How capable is this?

• 23:03Speaker 00

  How capable. But, I mean, that, seeing that does not inspire me to want to use AI more in my life. I mean, and I think it was last week I was talking about the fact that like, I'm not going to let an agent like get my credit card information or like get access to like, you know, to my hard drive, which obviously means that a huge swath of those capabilities are just not usable by me. And like a big part of that is because I just keep seeing stories like this. And I'm like, why would I let one of those things have, you know, like be able to make purchases for me?

• 23:29Kirsten Korosec?

  Well, okay, so I want everyone to mark their calendar. And this time next year, we're going to see how much has changed.

• 23:34Speaker 00

  Absolutely.

• 23:34Kirsten Korosec?

  I'm going to make a prediction that you will have given it access over the next year.

• 23:38Speaker 00

  Yeah, I'm not making this as a, this is not a forever promise. It's that there need, in the same way that there was a period where I was like, oh, I would never like make a major purchase on my smartphone. And then at a certain point I did, right? But I don't feel the need to be the, like, one of the early adopters here. Somebody else can do that with their credit card.

• 23:53Speaker 02

  Yeah, but Siri is not submitting a false police-

• 23:56Speaker 00

  So far, yeah.

• 23:57Kirsten Korosec?

  Well, I made a prediction, again, back to that sort of durability idea and like all these different startups that are popping up and getting substantial backing to tackle this issue of like rogue agents and security. My prediction is that a lot of these are going to be acquired, just sort of like what we saw in the first wave of SaaS startups getting swallowed up by like Salesforce. And what, what do you think is going to happen? Do you agree or do you think that some other-

• 24:23Speaker 00

  Yeah, I think there's going to be a lot of acquisitions. But I do think there's going to be value in independent companies that can like, I think trust is going to be really valuable over the next year or two. And, and it's going to be very hard for OpenAI or Anthropic or Meta to say, we fixed this, trust us, it's all good now. I think you're going to need someone, whether it's a different company, some sort of government agency, some sort of nonprofit to say, yeah, that's actually true. I think that if they just hoover everything up and just like, kind of just say, it's good, like, I don't think that's going to solve the issue, at least from a perception point of view. Probably not from a technology point of view either.

• 25:00Speaker 02

  Zuck's world is golden. You can take him at his word when he says, trust me. I don't know. I mean, I think, I think acquisitions probably are on the horizon. Also, as I'm broadly curious about this in AI, the tension between a company starting up its own take on security, you know, sort of agentic work, whatever it is, and the bigger companies just being able to create that internally. You know, these models are becoming so much more capable. We're hearing so much about companies being able to create very specific and bespoke tools using this technology. I'm very interested in that tension too. Like, how much of the momentum towards acquisition gets pushed back by the ability to just say, that's a great idea, I'm going to go build it myself.

• 25:45Kirsten Korosec?

  Yeah. Well, I'm curious to see how many deals happen here at Disrupt this week.

• 25:51Speaker 02

  Hey, if you want to back my Pinkerton AI agency, come see me after the taping.

• 25:55Speaker 00

  It is so depressing to think how much funding you might actually get for that.

• 25:58Speaker 02

  No, I hate it.

• 26:00Speaker 00

  I know, I'm not saying it's good, but I bet you could get somebody to write you a check for that.

• 26:05Kirsten Korosec?

  They may be sitting in the audience right now.

• 26:06Speaker 02

  The Pinkertons are actually backstage waiting to hit me with a copyright lawsuit.

• 26:10Kirsten Korosec?

  I am looking forward to a number of things this week. I always look forward to you introducing the Battlefield participants, by the way, because you bring such joy to the role. So I can't wait to watch you.

• 26:22Speaker 00

  Yeah, I'm excited.

• 26:24Kirsten Korosec?

  And then in just a couple of hours, I'll be interviewing the CEO and founder of Rivian. And I do think that everyone should watch that one because there's a lot to talk about. This is a company, speaking of AI, that started out as an EV maker and kind of is in robotics and building its own AI assistant and building an agentic system for its vehicles. There's a lot going on there.

• 26:49Speaker 02

  And a company that's probably older than a lot of the startup founders that are here. It was founded in 2009. We're getting there.

• 26:55Speaker 00

  Oh, that's depressing.

• 26:55Kirsten Korosec?

  Yeah, yeah.

• 26:56Speaker 00

  Wait, if we're previewing Disrupt, what's the deal with the space gun? Is somebody going to get fired into space?

• 27:01Speaker 02

  I do not think somebody's going to get fired into space. But I, you know, you can never rule anything out.

• 27:05Speaker 00

  Not going to.

• 27:07Kirsten Korosec?

  Gotcha. Well, is there anything else that you're looking forward to that folks should tune into?

• 27:11Speaker 02

  Oh, I think, I mean, I'm just so excited to get on the floor and see more of the Battlefield companies. It's, it's so exciting for us to be able to watch them come through the process and read about them on the page, but then see them in action and see, you know, the bolt robot like go bolt the bolts on, on the pipeline or, you know, the robot that's supposed, I guess I just like the robots. The robot is supposed to clean buildings. Like, it's so much fun to see people in person actually, you know, walk you through their dream.

• 27:39Kirsten Korosec?

  Yeah, absolutely. Well, we have to go, but you can always find us on social.

• 27:44Speaker 00

  At EquityPod on X and Threads. And for those of you who don't subscribe on your favorite podcast app, you can also check out Equity there. And we also are on YouTube, so you can watch us there.

• 27:54Kirsten Korosec?

  Yeah, thanks so much.

• 27:58Speaker 03

  Equity is produced by Teresa Locan Solo with editing by Kel. We'd also like to thank TechCrunch's audience development team. Thank you so much for listening and we'll talk to you next time.