Regulators Warn of New Era of Cyber Risk from AI

Bloomberg Audio Studios, Podcasts, radio news. Bloomberg Tech is alive from coast to coast with Caroline Hyde in New York and vow In sent Francisco.

This is Bloomberg Tech coming up.

Concerns keep rising around anthropics new AI model mythos as US officials won Wall Street that the tool could usher in an era of cyber risk.

Plus, we speak with Roblox CEO David Zooki as the company introduces special accounts for younger children and teens.

And we recap the historic trip around the Moon and back after the Artemis two cruise splashed down safely on Friday evening.

And ed on Friday evening, we were wrapping up a week for the stock market that was the best. That's November on the NaSTA one hundred, and some of that cautious risk on attitude still sits with us on the NAST one hundred today. As we see the oil is thrusting higher. Why then the optimism in markets when the inflatory pressure is clear, and that very sensitive view on

Ed.

Okay, let's get the latest details on Iran and the administration Frolomber Washington correspondent Tyler Kendall. Tyler go through the content of the president's true Social post that came just after ten am Eastern time and give us the latest of the situation, particularly with the straight of horror moves and with Iran.

Hey and well at this point, this new threat coming from President Trump aren't True Social saying that if any of Iran's attack ships come near the US IS blockade, then they will be quote immediately eliminated. This comes after a flurry of comments from Iranian officials this morning saying that the country is preparing to take action if it's shipping is threatened, including threatening to hit ports in the

Many a headline Tyler Kendall reads through it all for US, we so appreciate it. Look, with oil back above one hundred dollars, stocks while they've been wavering today still managing to cling onto gains. But President Trump is of course ordering that blockade of the straight Orf forward moves, and we want to give you the macro picture darkening, but also focus on some microews. Intel has managed to stage

What you see, Yeah, well this stock has had really an incredible run. So it's up today for the ninth day in a row. This is more than fifty percent gain in that time.

And it's had really.

A spate of good news that's just gotten investor momentum back into shares.

Here.

So you had the news that they're buying back half of a plan in Ireland from a public global management. Then you go into the announcement that they're going to be with Terrifab, you know, to develop semiconductors for Tesla SpaceX and then they also announced a commitment for Alphabet's Google to use future generations of some of their processors in data centers. So these are all great things that investors are really looking for. They want more coming out

And it's data that I had to double check on the Bloomberg Tunnel, not because.

I don't believe you. I do.

It's unbelievable of itself, right, that level of gain over nine sessions, fifty three percent gain. But you make the point just building what you just said that actually last year was pretty good as well, and so there's a broader recovery story with Intel. What is the market saying about how they feel on the turnaround?

Yeah, I mean, I think investors are really voting with their dollars here. They're feeling good, they're feeling like there's a progress being made, and they're looking really long term for this company going forward. I mean, I think it's important to point out that it does report results in two weeks, but investors are looking even more long term for progress in their foundry. They want to see more customers there and seeing that, you know, there really could

Isn't it funny that it was Jay Goldberg of Seaport who gave you that view, and he's the only cell rating really everyone in video, so that sort of supports is his bearish take on that company as well as bullish on Intel. But the rest of the ANAS community are a bit more mixed. So you paint a picture of this stock getting expensive.

They yes, So the evaluation is actually quite high. It's trading let me say why I get this right. More than ninety times estimated earnings over the next twelve months. That's very high, especially when you think about the rest of sort of the semi space, and you know in Vidia, I think in Video is at about twenty one times forward earnings. The thing that I hear though from investors in analysts is that maybe it's not the right way

And for what it's worth.

Intel up almost three percent in the session, ninth straight day of gains, longest streak of gains since September of twenty twenty three. Bloombers come and Ryani Key with the must read on the Bloomberg terminon dot com. Now coming up a conversation with Caris Fragg Hacker one CEO and Peter Singlehurst to Bailey Gifford on Anthropics' latest model mythos and in part it's soaring valuation that's next.

This is Bloomberg Tech.

Wall Street banks including JP Morgan, Goldman, Sachs, and Morgan Stanley are part of a small closed group with access to anthropics Mythos. They're testing the powerful new AI model to identify vulnerabilities and strengthen defenses. After Bloomberg reported the US officials warned Wall Street the myphos kudosher in a new era of cyber risk. Other global regulators are watching closely.

Joining us to.

Discuss is Cars Spragg's CEO of the cybersecurity firm Hacker one. And there is a great reason to have you onto the program to talk about mythos and the cybersecurity context because what Hacker one does, it's the biggest offer of bug bounties.

Right.

You pair companies all sorts with someone that can identify vulnerability and then in some cases suggests how to patch it. That's exactly how the banks are supposed to be using methos. And I wonder just initially your take on the model and what you know about it.

Well, I think it's very exciting that we are starting to see capabilities where we can at scale identify vulnerabilities more quickly and use them in a defensive capability to eliminate that risk, and so this is a very great compliment to what we have in doing for decades, which is human driven vulnerability discovery. Where the bottleneck truly is though these days is no longer really with the vulnerability discovery, but it's much more in the back part of the

Kara.

Last week I spoke to Mike Krieger, who co leads Anthropics Labs, and I tried to get him to just be succinct about why mythos is good in a cybersecurity context.

Listen to what he had to say.

You can't isolate a capability. I mean you probably could, you know, with enough effort, but that's typically these things kind of emerge. The fact that it is very good at solving general purpose coding problems and debugging and doing all the things that you'd want it to do in the ordinary course of practice also makes it really good

If you can't isolate a model single capability in this case cyber, why is it so good in that domain?

Well, I think about it as coding and building is one side of a coin and breaking is the other side, and they kind of go hand in hand.

And one of the truly.

Big advancements that Mythos has is this ability to chain together exploits and basically turn what could be multiple vulnerabilities into a much more critical or severe issue. And that is fundamentally building and putting something together across a number of building blocks.

Can you, Kara just push back against pats some cynicism that's been building into the market about the power of Mythos, some saying look at benefits anthropic to make out that it's quite so powerful and actually it's not doing all that much new than was already out there, or indeed it's just more compute issue than actual shared terror. As to why they put it into the hands of a small amount of players articulate just how powerful this is.

Well, I think the capabilities that are explained in the Mythos release are are quite good, and they're definitely in an advancement. The frontiers is certainly moving, and yes, some of those capabilities can be achieved with other models out there, perhaps through a more complicated workflow. But what we're seeing here is a general advancement and the ability of AI to really play a powerful force in the cyber security space.

Up, and that frontier is being moved by other players as well, and we know the opening. I's been trying to work with a cyber community as well. Just we had Fortealiss join the show last week and their idea is, look, smaller companies should be able to have access to Methos to be able to drive forward this vision to make sure we're safe. Is that something hacker one wants? Would you like to be on this small amount of group through without their testing vulnerabilities?

Well, certainly we would like to have access to this.

Now.

I recognize the risk equation that the Anthropic has taken here in terms of a smaller release to a limited number of organizations, but look at Hacker one, for example, last week alone, we automatically validated four thousand vulnerabilities that came in through our various vulnerability discovery programs. And that kind of capability and Mythos that would help us supercharge our validation capabilities would be greatly appreciated.

The access to Mythos is clearly being closely controlled. It seems strange to not bring up the war in Iran and the competence that Iran has in as a cyber bad actor. You know, put two and two together. What is the likelihood that the malicious actor from Iran or elsewhere would have access to that level of technology and pose a threat.

Therefore, I'll put it this way, ed, I think this year, in twenty twenty six, civilians, businesses, and organizations are markedly less safe than we were just even last year. From a cybersecurity perspective.

We now have much more.

Capable AI models. Those models are rapidly proliferating, even if Mythos itself is still within limited release, and we now have a number of sophisticated thread actors that can put those capabilities to use, and we're seeing an increasing number of breaches and supply chain issues across the open source ecosystem and across corporations.

Kara, it's been great having your expertise today. Thank you, Krstig, CEO of Hackle One. Now let's stick with Anthropic It's power and also as investors way the powerful private market valuations across all of AI. Let's bring in Peter Singlehurst's head of private companies at Bailey Gifford, and you have,

I think when you own a business and when you're investor in any company, you always have concerns around companies. It's hard job to worry about the holdings that we have. But I think that what we're seeing at Anthropic at

What has been notable apart from Mythos has been the adoption the arr that they're managing to post thirty billion dollars and the amount of enterprises that are now using it, even in the face of concern that they are well sewing the government because of the current blacklisting single out by the Pentagon. How is that narrative driving what you think about Anthropic at the moment.

When we invested in Anthropic, one of the reasons we liked the business was the principal approach they took to safety now that I think has informed how they're approaching the Mythos models, and it's informed the approach that they've had with the Department of War.

So whilst we would of course rather not.

See them in this disagreement with the Department of War, I think and the company have said that there's more that unites them than divides them with the Department of War, and I think it's important to us to see a company standing by the underlying principles which enables them to flourish and enables them to attract and retain key talent.

What has been so interesting about the headline generation from Mythos is that people are looking at the model itself. You know, Caroline's absolutely right, the revenue run rate, all those kind of classic measures for late stage startups, but just your sort of response to the model's capability. Anthropics place in a group of five or six frontier labs that are trying to do exactly that very big model with great capability.

Yeah.

I mean, what we've seen with these lms is that the capabilities of the leading edge models will quickly become available to the trailing edge models. And then what separates the likes of Anthropic or open AI is their ability to continuously stay at the leading edge. And so, as you know, as a citizen, it does concern me that there will be these capabilities which today are in the

A lot of focus in the United States on anthropic There was a headline this morning that open AI, which you are not an investor in, is open a a permanent London office having wound down the Stargate project in the UK. I ask you all that basically to push on whether people outside of the United States see that same level of battle that's going on between the frontier labs, and whether the release of a model like Netho's impacts everyday people in the United Kingdom or in Europe.

I think in the very short term, the everyday person on the street is not impacted this week or next week. But are these models going to impact all of our lives in the months and years to come? Like yes, absolutely. I think that here in the UK, those that are in the industry, those that follow this are just as outware and on top of what's going on as his

I mean, Peter, what hasn't been notable is while we're all spell bound by AI, the whole world has been spell bound by space over the course of the weekend. And I'm interested, therefore in your exposure to SpaceX and exposure to XAI in many ways, how are you thinking about this moment that everyone's world changes if they get access to some of these big companies not just being private, but potential to be public as well.

I think it's important that every day investors are able to access these kinds of companies, and so I think these companies coming to the public markets is overall a good thing. I think what is in many ways a shame for the everyday investor is that there is so much growth that has already happened before these companies come to the public markets. And it's not just anthropic. It's not just SpaceX. You could look at Byte Dance, for instance,

Peter, I know we've had the opportunity to speak to you a lot recently, just very quickly. At this point, does the market understand the economics of orbital data center ahead of this space X ipo.

No, I don't think the market does understand the economics of orbital data centers. I think that this is a very nascent idea in Nason technology, and where that value can potentially accrue I think is through the vertical integration with the models. If we get to this place where power for computing AI models is existential pinch point for these businesses. So in short, no, I think it's not yet well understood. It doesn't in a sense exist yet.

Peter Singlehurst of Bailey Gifford, thank you very much. Now coming up, a Trump family crypto venture. It's facing some investor backlash. We got more on that next, this is Bloomberg Tech.

It's time now for Talking Tech and first up, World Liberty Financial. It's a Trump family crypto venture and it's facing an investor revolt that includes billionaire backer Justin's Sun, who says the project is secretly building controls that len insiders, freeze tokenholders funds and extract cash before a wave of tokens are then unlocked.

Look.

World Liberty has roundly rejected the criticism the token trays.

That are out there.

It's very specific data, but it speaks to one clear thing, which is AI is shrinking the market for white collar entry level jobs. Data input, basic coding are the graduate level jobs that.

You would bank on.

Now those people that have degrees can't get them. So they're going into restaurant, retail service industries to find employment.

Yeah, even babysitting. Look basically, they've got to make the money, still work for them, But what is the job of the future. And there's a lot of anxiety in those that are still at university. I speaking with the NYU Stern professor Run Sundarajan just last week and him saying the amount of anxiety among his current tuition folks.

But they're also all.

Getting more entrepreneurial. Everyone's spinning up a business set right and.

At the same time college applications going like that.

In Congress, we're coming up.

We're going to speak with the Roblock CEO Dave Bazuki as the company announces new age based accounts and experiented parental controls. This is Bloomberg Tech. Welcome back to Bloomberg Tech. It's still the war in Iran and the situation in the Middle East that the big driver in markets.

At the index.

Technology stocks modestly higher basically flat er up two tens fve percent and then as that one hundred. But within that index the outperformances in software names. Remember last week we saw some pressure on software names that all kind of related to displacement from AI. But the other big factories also oil Brent the global benchmark for oil. You're above one hundred dollars a barrel again on Brent, and the situation is a blockade in the straightfor moves who went through it earlier.

In the program.

There isn't a lot of newsflow outside of that carrier that's impacting the index level for tech, but there is other news that's driving single names.

There is and let's check in on Roadblocks, which are significantly in the green today the online gaming platform introducing new age based accounts, expanded parental controls for users under sixteen, and it's of course the latest effort by the company to really enhanced safety for its younger audience. The update comes as a company prepares to roll out also a new subscription plan later this month. Got a lot to

Hey, thanks for having me on.

And you know, when we've talked in the past, we've shared our commitment to setting what we believe is the global gold standard for healthy, safe, age appropriate digital engagement.

We've been innovating on this for years and years.

We started with age check in January for communication.

What we're rolling out today is the next phase.

In that implementation of that global gold standard. That of course includes all the other things we've done in the past, including innovating on text filtering and non sharing of images. So today we're announcing kids accounts and select accounts. We use age estimation to automatically assign people to the right account and we connect them with appropriate content. For kids' accounts, they get mild content, things like grow a garden.

For our older users.

In select accounts, they get moderate content as well, things like emergency response. And users who are over sixteen of course get regular roadblocks accounts. So this for us is just a continuous evolution in our commitment to this establishment of a global gold standard for safety and healthy digital engagement.

Let's talk about the underlying technology and as you literate, it gets better and better. But age verification is very tough.

How good has it been?

How many people's age is it getting right? How many people push back or try to dispute.

Yeah, we started experimenting with facial age estimation middle of last year laid of last year. We saw that it was more than good enough to estimate age, and in addition to that we use additional signals. We're constantly incorporating other signals in addition to facial age estimation to put

Dave, the balance is between sort of codifying, taking action on age verification, which you've done, and then the impact on engagement daily active users. What have you and the team modeled for on how you see daily active use as an engagement responding to this new set of rules.

Yeah, yeah, thank you for highlighting this, and I do want to highlight as we run this company, we're always leaning in on doing what we believe is the right thing for our huge community almost one hundred and fifty million daily active users, and respecting that community and building an innovative product with safety built in by default. We've seen sixty five percent of our US daily users now age check. That number continues to grow and we're optimistic about it.

So we have modeled this in.

We have a lot of other things coming throughout the year as well, and we continue to focus on doing the right thing for our community.

How much does weather A game is addictive or not factor into this, Dave, And as you know, the broader question is whether games gaming platform should be talked about in the same conversation as what we're seeing in social media bands around the world and the conversation on social media addiction.

I want to speak out for the gaming industry as a whole relative to social networking and social media. There's something really important about multiplayer gaming and experience. It's a connection activity rather than a being alone activity. It's an activity that typically involves creation and imagination rather than raw consumption.

With a safer environment not only comes well, parents and children more willing to use, probably advertisers even more willing to come on. And that's an area that you've been developing, and particularly a revenue share with ads that are going to be based within developers' games. How is that being taken on by the developer that you work with, and how you think about the revenue share? What sort of amount to me thinking, yeah.

Yeah, I want to highlight once again, we always put building a healthy, safe and age appropriate environment first amongst everything and once again respecting that community.

But as you.

Correctly mentioned, in immersive three D environments, if I go to a NASCAR race, for example, I may see some interesting products on the cars or on the billboards, and those are typically things that are appropriate for people of all ages. So we continue to build out that advertising opportunity on roadblocks. In addition to those immersive types of ads, more and more of our creators are actually buying ads to help people recognize that they have a new creation.

So we're still early in.

The advertising opportunity, but we believe it's very large.

How large could the revenue split feed? Have you got an idea of where that percentage WI ledge out on particular developers.

Well, we do have kind of internal models and we're very optimistic that. You know, for creators building games and experiences on roadblocks, they primarily monetize right now with virtual currency, with our Robucks system. There are millions and you know creators on our platform, and a portion of them actually

Dave.

I'm looking at your company shares up almost five percent in the session. I'm trying to work out what the market's really responding to here and Friday Night, Roadblocks launched. This new subscription plan goes live April thirtieth, I think, but it basically replaces the premium tier. There seems to be a lot of good vibes around that. What was the rational and how do you see it playing out?

More and more we see two types of subscriptions on our platform. A large number of parents, you know, the future of what we see happening on Roadblocks. When I was younger, I got an allowance to go buy comic books at the local drug store. Today, a lot of young people are learning about digital finance by getting a Roebucks allowance from their parents. So we do have a subscription for Roebucks so parents can give their kids recurring

Dave a zookie of Roadblocks, important pieces, several piece of news out of the company, Thank you very much. There's some interesting market activity that's going on. Look at shares of Dell and HP Inc. Both of them moved very quickly to session highs after a report from semi Accurate, a website which basically said that Nvidia has been in negotiations for every year to buy a large company quote that will reshape the PC landscape, without naming the specifics of it.

You know.

We note in a BFW on the terminal that semi Accurate has been reporting on this kind of ongoing story since twenty twenty four. Late twenty four, but the stocks both moved higher. Dell's up four and a half percent as it stands, HPN cup two percent. Bloomberg News is out to comment for all three of those companies and yet to hear Backgary.

Keep an eye on it. Meanwhile, coming up that record rating Artemis two crew back on Earth, NASA and its partners turned their attention now. Of course, they're landing humans on the Moon as soon as twenty twenty eight. We'll discuss that next as bloom meb tech.

After blasting off from Earth the historic trip around the Moon and back, the Artemis crew splashed down safely on

NASA administrator Jared.

Eisaman posting that the discolouration scene in pictures after Orion was recovered was not parts of the heat shield that have come loose or unexpected behavior in the materials as it happened in Artemis one. NASA is now doing a full data review of the systems along with his partners, including Lockheed Martin, the company's head of human space exploration, and the URON program manager. Kerk Shiman joins us now

Absolutely, the vehicle performed exceptionally well, we're very, very pleased with how it performed. The heat shield did exactly what it was supposed to do. It protected the crew and the vehicle from those very high temperatures that you mentioned.

I actually got to see some pictures that.

The divers, the Navy divers took from underwater, and they looked all exceptionally Well. You know, when you come in and you have that high heating and then the atmosphere, the wind and so on, you expect to have some discolouration from that experience. But all in all, the vehicle performed the heat shield and the tiles on the top of the vehicle performed exceptionally.

Well, we're very, very pleased.

We're looking forward to getting more detailed photographs right as we have the vehicle back in San Diego and on to Euston.

Hey, Kirk, that's just good old fashioned rocket science for a flat bottom capsule, right, how do you slow it down on the re entry interface?

Bang it into a wall of air?

I mean, on this program we talked about how to fix Atomis one to Automus two.

Is actually the angle of.

Re entry rather than some kind of fundamental change to the materials of design.

But with Automis three.

There will be some tweaks, right, what can you tell us about that?

Sure?

On Arms three we use the same shape of av coat tiles, but the formulation of av code is slightly different, and so we're looking forward to different performance, actually better performance, and we can basically skip out of the atmosphere shortly and extend our downrange, which is really important for our operational constraints. So looking forward to a more operationally friendly heats shield on Artists three and subsequent vehicles.

I mean, Kirk, there must have been lots you're looking forward to and getting your hands on it back in Houston, and I'm interested in the early day and more broadly the crude feedback. What is the most important things that you've learned from integrity?

Well, of course everyone.

You can imagine how we engineers are interested in every little detail that we learned along the way.

I think the beautiful thing about.

Having the crew back and talking about it is kind of what we've learned about our planet and where we sit in the universe. To me, it was so exciting to hear them talk and basically see the Moon and ultimately our Earth through their eyes.

It was just fantastic.

Hopefully you've had a chance to hear that, and if not, you will over the coming days.

It's a joy to see so many almost get philosophical in these moments when it comes to planetary exploration and the moon and just how we sit within this universe, and you too. But when you are thinking about how the engineers get nitty gritty and some of the key things that you're looking to confirm before Ryan flies again, what do you think is the most detail oriented focus you're going to have.

Sure, I think everything that didn't perform exactly as expected on this flight, or where we'll really dig in. Lockheed and NASA spend all their time or ninety percent of their time looking at the things that didn't go right as opposed to all the things that did.

They was one thing that you're writing about one other things that perhaps didn't form exactly as you anticipated, even if not badly.

Sure, we had a valve that didn't work right relative to the water, the drinkable water, the potable water system. We had a system that does introduces helium into the propellant tanks as we deplete propellant that didn't work exactly as expected.

So we'll definitely dive into those issues.

And then a few little features with telemetry that we'll have to fully understand. We think that's probably aware issue, but we'll understand it and correct it for future flights.

Christina Kirk mission specialist turned astro plumber.

I don't know how you would describe that.

You know, I appreciate the answer of the things that didn't quite go right, and you know there's a data process, a review process, right, but what exceeded expectations. You know, everyone makes a lot of the idea that integrity or ryan is a spacecraft is the size of a small bus, a Campa van three.

Hundred and thirty cubic feet, right, you know.

But it got them two hundred and fifty thousand miles plus from Earth, around the Moon and back, and by all accounts, they had a great time.

What exceeded your expectations, Kirk?

Sure, Ryan has twelve million, five hundred thousand parts, so most of those parts did exactly what they were supposed to do. I think the big surprise, or not a surprise, but exceedance.

Of expectation, was the systems that keep.

The crew alive and happy, that the scrub carbon dioxide out of the atmosphere, that introduced auction in the atmosphere. That basically all those systems that keep the crew alive and healthy and then interfaced with the crew to the vehicle performed flawlessly.

It was just exceptional. And that was all new for this flight.

So we were expecting to learn things, but by and large those exceeded our expectations.

Cok.

This was a great feat for NASA and for America. But we should be kind to your European colleagues, the European Space Agency and the companies involved in the service. Module absolutely critical for life support and other critical systems. How did the service module perform, To the best of Lockheed's knowledge.

It performed exceptionally well.

We have a great relationship with Airbus, which is the industry partner in Europe that built the service module, and the European Space Agency. They were there in the Michigan control room sitting right next to us during the entire mission, including splash down, and again they were extremely thrilled by the performance of service.

Modules, as were we.

Of course, there were one or two things that didn't work exactly as planned, and we'll get to the bottom of those with Buy and Large. It performed better than we better than we expected it to form.

It was exceptional well.

I think the rest of the world was highly pleased. It went so well as well. KERK Shaman, thanks for giving us your time. VP of Human Space Exploration and oryan program manager for Lockie Martin. There now coming up.

Apple moves to.

Decentralize its AI leadership as it is is it's next hardware hit on that next this is Blue Medtech.

Apple's hitting a major inflection point in its AI journey. Apple's AI head is officially departing this week, while a new product roadmap is coming into focus. It's all detailed in the late this power on by Bloomberg Smart gunman who joins us now display free smart glasses.

What do we know? Give us a little bit of the history and really.

Yes, really smart glasses. That is Apple's next big thing. They've grown in popularity right. Tons of people are wearing them, not only in Silicon Valley but in all parts of the world. Meta has done a great job with their smart glasses. Here, I'll put them on for you. These are the Meta ray Ban display glasses. So these have a little heads up display in them. Apples won't, like you said, display free. The idea is that this is the eyes and ears for AI and the cloud for

Maybe we'll all look a bit more like Tim Cook as you say, one of the slimmer rectangular designs is sort of based on what he wears.

Mark.

But the age old idea is that Apple comes later but wins. How much do they need to win? How much does that need to be integrated with the new leadership in AI quickly?

Yeah, well, smart glasses story for Apple is going to be strong. I think the big leg up Apple has is this thing. The metaglasses don't pair as nicely as Meta would like with the iPhone. Any third parties for our classes don't pair as nicely with the iPhone as Apple in house Glass as well, and so given them a number of iPhones in use, the attachment rate I think will be material for the company. So I think they're very much looking forward to getting these out the door next year.

Mark German as always with the power on over the weekend, which remains one of the most read stories we can week out. We appreciate it.

Now, that does it.

For the edition of Bloomberg Tech Ed, we did a lot of AI, a lot of space, a lot to digest.

Yeah, maybe not as many breaking US headlines technology stories, but a lot of deeper conversation about things have been going a while now. Check out the podcast to recap all of that. You can find it on the terminal as well as online on Apple, Spotify, and iHeart This is boombag Tech