AI Training Data Secrets, Stealth Models and Account Control

Hey everyone, happy Monday. It's Monday again. You know what happens every day after Sunday.

No matter what, it's Monday, and here we are on Monday again for Techstrong gang. We've got what's forming up to be our kind of Monday regulars, and we're happy to have them here. Let me introduce you to them.

First of all, I want to introduce the birthday boy. My friend Pawel Plawitz is celebrating today, and I can't think of a better way to spend your birthday than on here talking tech on the gang. Pawel, thanks for joining us, and happy birthday.

My pleasure. Thank you. Joining Pawel, some of our regulars, the one and only Yvette Schmitter.

Good to be here. Hey gang. The two and only.

I don't know, what do you say after one and only? The two and only? No.

My friend Stephen Foskett. Good to be here. Yep.

And the master of the guitars, Mitch Ashley is with us, and of course, sitting up in the high castle, Mike Vizard. Hello. Gang, welcome to Monday.

I hope you all had a great weekend. It was, as usual, not a quiet weekend from the news perspective, but we're looking at a couple of, well, usual three blocks of stories today. Mike, we got a bit of an AI whodunit mystery.

And you're right about the weekend. It was kind of crazy. We're probably set for things to talk about through Wednesday, but let's jump into this first one.

Yeah. We have a mysterious new model showed up called Ox Alpha that was apparently traced back to China, and it's reportedly very powerful and also comes with some free coding tools. And, it's oddly enough, the president took this opportunity over the weekend to also tell people that they should get behind data centers because we're losing this race to China.

" Or what's your thoughts here? As a person from Europe, I should say that the race is not over, right? I don't think the race is over.

The race already started and started to be interesting, really. So yes, we have two dominant sides, the US side and China side. And this model is exactly as you said, is back to China.

So it rises a couple of questions, especially that they don't want to say anything about this model, and they just claim that no data is stored, and more you claim this, more I don't believe this. So that's the one thing. The model is quite powerful, at least from the description of it.

So it's not very huge model, only 1 million token context window. However, it's free or mostly free, so this is the big power of this model. " But let's try something responsibly, to be honest.

So, as we know so far, it's pointed to the Xiaomi team at this point, so generally well-known company. However, again, as a Chinese model, it has its flavors, I would say. Depends on the country and the region from where you are asking things.

And, I would say still, yes, be cautious using it. As long as we don't know to whom this belongs, what is the idea behind the model. Mitch, I'd love to get your thoughts here because I actually went to a dinner last week with the folks from Snowflake, and they've added Chinese models to their systems, and you can use them happily.

And they say that they've applied all the governance things, so you don't need to worry about where the model came from, and they'll take care of that. So from your perspective, these Chinese models, are they safe to use, or what should we be concerned about? I think it's becoming more acceptable to use the Chinese models, not everywhere, but within some controls.

You're not going to use that on a federal government project, obviously. Maybe, probably not even in a major enterprise environment, at least not yet. I would be very cautious about it still, but there are other applications for it.

" I know the security person in me wants to go out and try every model that shows up with no attribution of who created it. So it either was a way to grab attention, they weren't proud enough of how good it was, or they just don't want us to know who it is. Who knows what the real reason behind that is.

So, I would be super, super, super cautious, and let folks who want to experiment with it in a controlled, locked down environment. It's got development tools. Sure, you could do some things with it.

I guess it's going through the router, so I'd be really careful what I'm sending to it. Yvette, one of the things that occurs to me at least is that there's a lot more data scientists in China, there's a lot more data, and there's a lot more power. So, how are we supposed to win this race at the end of the day?

Because it seems like the deck is a little bit stacked on that side. " This is exactly what you do, and let me just tell you this. You and I spoke about this a couple of months ago where we were talking about DeepSeek and read the terms of service because they tell you exactly how they use your data, and even if you break up, you're still together kind of stuff.

" I'm a New Yorker. There ain't no free stuff. There's no free meal, there's no free ride, there's no free nothing.

So when someone is telling you that it's free, it ain't free because it's always going to cost you something in the end. And this whole thing, the reason why, in my opinion, we don't have rules and regulations and guardrails around AI is because we need to beat China. I'm going to say this, we've already lost in the sense of if everything to make us first is tied to capitalization around seven companies.

" And let me tell you, the buck is going to follow free. So in my perspective, I believe the war is done and dusted. We're behind, and if you want to go ahead and use that model, I'd say read the terms of service, and if someone tells you that it's free, I got two bridges in Brooklyn that I can sell you.

There you go. Steven, one of the things that's also come out that we haven't quite gotten to yet coverage-wise, but it's on the list, is that NVIDIA is talking about building an OpenAI ecosystem. And as part of that, it seems like they've read the tea leaves, and they're basically saying that the best way for us to respond is to invest in our own open source technologies, and away we go.

What do you think? I like it. I like it a lot.

We should point out first, open space AI ecosystem, not an opening AI ecosystem. That's a different thing. Yes.

Exactly. They have suggested that we need to have a proliferation of open weights and yes, open source models. And I hope that folks who are listening to this realize there's a difference between open weights and open source and free.

Those are three different categories of things, and sometimes they overlap and sometimes they don't. Yeah, honestly, think about what NVIDIA sells. Where do they make their money?

Well, of course, they want there to be more models, more use of AI. If open source and open weights and free providers are the ways to get there, then they would be all for it because everything helps their bottom line. And, frankly, I think it helps us too, because if there are more options, if you can run things locally instead of putting them in the cloud, instead of putting your trust in a big American company or a Chinese company or any kind of company, then I think that it's good for us.

And, I think that it's important to kind of think about things in these ways because, for example, with regard to this mystery model There's always a lot of FUD around, "Oh, you're driving a Chinese car. " Well, an AI model is different, especially AI service, that you don't know what it is because you're literally giving it your data. It's not like a mystery.

It's not like, oh, they embedded a grain-of-rice-sized monitoring module on the motherboard, which is what Bloomberg accused Supermicro of doing. No. " They got your data.

We know they got your data. So it's a different story entirely, and I wish that we would think about this. And hopefully, if there is a proliferation of open weights models that you could run locally, then nobody's going to get your data.

And I think that's smart. But the problem is, of course, it takes a lot of hardware, and that requires a big check to Nvidia, and Nvidia's happy, as you pointed out. Yeah.

Here are the thing, Mike and Steven, Mike, I'm going to go back onto my soapbox. People aren't reading terms of service today. They want to have the ability- They put next, next.

Right. They do next. They're not reading terms.

They have no idea what's happening today with their data because they want to create a video that makes a squirrel fly a plane and drive a yacht. We- Oh my God, can we do that? Yes, we can.

Oh, I'll talk to you guys later. This is where we are today. This is the world we're living in.

And I've said this before on this show, when your data is more valuable than your vote, you have to sit up and pay attention because the free here ain't free, because the free here gets paid in data on service you will never, ever see. So free is going to cost you. And running stuff, I was going to say and I'm going to say it, but I didn't.

Running stuff free locally, Steven, it's like, yeah, if you don't have an agent tied up to your web browser with complete access to your lap-- They're going to get your data if you're doing stupid stuff. Right. Alan, I got- So let me go on this, Mike.

I let everyone have their say. I got two words for you: free beer. Yep.

There is no free beer. There's no free anything. There's nothing I'll give up for the paper.

There's no free anything ever, as my friend Yugad says. But listen to me here. Who's sitting here saying, I would much rather trust slippery Sam Altman with my data than I would some Chinese open source model?

You're separating levels of hell in Dante's Inferno here. Heaven level. Okay?

None of these people deserve our data. But that being said- Oh, come on, Alan. Elon Musk is perfectly trustworthy.

Yes. And they've proven it time and time again. " I call it blind.

I don't care. That's exactly what this is. " We've got reservoirs at record lows.

We're not sure we've got enough water for people. We don't want 21 guests, natural gas turbines running off the grid next door and exempt from what little environmental regulations we have anymore anyway. Right?

This whole goddamn thing reminds me of something out of "Blade Runner," where we're going to exist in some polluted, abysmal land with neon lights written in Chinese languages, and we're going to have robots running around and everything else. So wait, I- I have to point out, "Blade Runner," that was Japanese, not Chinese. Okay.

I'm sorry. Excuse me, Steven. I stand corrected.

I read some of those articles, and they did point out that we are consuming more water on our golf courses than all the data center folks out there. So do we just need to stop playing golf? Is that what that'll take care of?

Well, no, because the golf center is being pushed by the Japanese, not by the Chinese, I think. No. You know what?

There is that thing, so Mike, that's an urban myth because that's looking at what data centers are doing now, not $8 trillion later when they build these out. Okay? So that's a red herring of a number.

But hear me out. The point here is, though, what does data centers have to do with this mystery? And then Steven, you're right, open weight, not open source.

What does it have to do with this new AI model? Right? Clearly, it's Chinese.

Clearly, they did this for a reason. Clearly, they did a great job of marketing there. They went open weight to combat the American Cadillacs.

They came out with a bunch of Toyotas. Right? And there's a lot more Toyotas than there are Cadillacs in the world.

Right. And we have no one to blame but ourselves. Let me ask you- Why can't we produce Fords and Chevys?

Let me ask Yvette this question. Is that quiet sound I'm hearing, is that IPOs being postponed yet again? What do you think the ultimate impact here is?

That train has left the station. I don't think those IPOs are going to be slowed down or delayed at all. " Right.

You have all these movies. And I'm literally saying our current administration has rolled back, destroyed any type of levels of laws and regulations around the EPA. That's why Elon can stick those turbines down there.

That can put safeguards around our environment to make way for these huge data centers so that you could do, create more video of the squirrels flying planes and driving yachts. So I do think that we as people, Mike, I'm going back on my soapbox, we need to understand that right now is right now, and it won't last forever. And when you look up on your tiny little screen, the world is going to be completely different.

And the opportunity that you had, you gave it away because you wanted to create a squirrel video. If the Chinese- When Greg Abbott- Yeah ... starts investigating data center licenses, that's a statement.

I think if the Chinese had a sense of humor, the next one of these would be called the Manchurian Candidate. All right. On that, you know what?

Stick to the day job. Let's hop to our next segment, though. We're a little over on this one.

All right. Books, and it's another mystery. But apparently, it started out in Europe, I believe, but there was reports where large amounts of used books were being shipped to places unknown until somebody put a little one of those Apple tracker things into a box, and it showed up it was going to an Amazon data center.

And apparently, the issue is that, while if you download something, you might be violating copyright, but if you grab a physical copy of a book and then scan it and then shred the book, you seem to be okay within the confines of using established content to train an AI model. Steven, I imagine it's legal, but it doesn't feel like it's a good look for the AI guys one more time. It sounds like it's legal.

But yeah, first kudos to the 404 Media people for doing this. I love the idea. Rare book seller gets a big order, 404 Media throws an Air Tag in there and watches where it goes, and it goes to an Amazon facility in Las Vegas.

Also, kudos to the self-aware Amazon faculty or facility employees for creating a really on-point logo of a dinosaur destroying a book. I couldn't have done it better, honestly. And honestly, it looks AI generated, so double kudos for that.

But the point is, as we've seen, as you talked about, essentially, Anthropic got in big trouble for illegally downloading thousands, millions of books, or images of book content, and using that to train their models. 5 billion, which was a huge fine. And authors, including some of my friends, actually got checks, $3,000 checks in some cases, for a book that was ripped off by Anthropic.

" And so they're buying all these books and, yeah, ripping them and scanning them, and then destroying the book, because that is what this judge said is legal. But to me, I think the problem is it's a case of a law not keeping up with modern use cases. Because we're conflating a person reading a book and learning from that book, and maybe even emulating the author of that book, and an AI reading all the books.

That's a very different thing to me. But I feel like, in this case, the judge didn't see that nuance and felt like AI reading all the books is the same as person reading one book, and therefore, the law is clear. But what do you guys think?

Is that the same thing, or am I completely off track here? No. No.

There's a nuance that, Stephen, I love your insight, but we missed a piece. So old print is the target. Old print.

Because it predates any of the AI text slot created flood. So it reads as human clean data. That's one thing.

And the legal fault line, we need to unpack that, because the judge blessed scanning lawfully bought books as fair game. So while the very same act may already be viewed illegal across the EU, in the US, where do those books end up? Las Vegas.

A Las Vegas, Amazon warehouse. So I do think that folks need to... 5 billion for the fines, but a $3,000 check for my life's work in perpetuity, for me, I'm worth more than $3,000.

I say there's so many- It probably went to the publisher, by the way. It usually doesn't go to the author. It usually goes to the publisher who owns the rights in the first place.

There's so many connections to things I can think of, like during every apocalypse, doesn't everything end up in Las Vegas? Number one. Two, this sounds like a mash-up of Ray Bradbury and "Fahrenheit 451" mash-up with Napster.

We're kind of reliving history over and over again. And expecting the courts to understand the nuance of this, they have a tough enough job in more or less things that are really- Clearly, Mitch, the judge created a loophole here. Yes.

That was not his intention. That was not the intention, but that's what other lawyers do, is find loopholes until somebody else closes them. And sometimes we're in favor of those loopholes.

I just want to point that out. My local library lets me borrow e-books on a one-for-one basis. So basically they buy one, and then I can borrow it from them and read it on my Kindle, and then I can return it.

Now I'm not actually borrowing or returning anything. I'm doing that digitally. And it's the same with the EFF.

Or, I'm sorry, not the EFF, the Internet Archive. They have bought and scanned books, and they have a library of books that you can borrow from them, and that is perfectly legal, and I feel like I want to be in favor of that. But is it really any different between Anthropic doing that and the Internet Archive doing that?

I say yes, but only because I'm looking at the nuance. So Pawel, I haven't heard from you in a bit, but let me ask you this. Yeah.

Are the Europeans looking at this and, by the way, are all the bookshelves and the used bookstores in European capitals empty? What's going on? So yeah, I would say we don't like it.

There are a couple of things, really. You mentioned, Stephen, about borrowing like a PDF or whatever. You borrow a book.

Here, the book is no more. It's scrapped. Can you access this book through OpenAI or Anthropic or whatever?

This is the one thing. Second, to add to this stuff related to pollution from data centers and so on, we print books and just scrap them. Very good choice.

And also what the most amused me in all of this really, those companies are okay, doesn't matter if there's a loophole or something, whatever. They're okay to, let's say, borrow the work of others, like you even mentioned, and that's okay. It's for community, for humanity.

But then they are fighting each other that your model is too close to mine. There is some legal issue for sure with that. So I feel like being in some crazy movie, let's say, and I understand what's going on, but I cannot find why.

Because I can't find any real standpoint why those things are happening in that way except the greed. And I would like to think about humanity, that we are a little bit better than that. So this too shall pass.

I had this conversation on LinkedIn this morning with someone. What seems to us right now to be the way it's going to be is not. Because the fact of the matter is, this is exactly the problem with LLMs.

They need a new drug constantly. If you want to get these LLMs to AGI, super intelligence, whatever, better than they are today, you've got to constantly give them fresh data. It's like heroin.

We got to feed the new... I got to feed the drug. And we're running out of fresh data.

We're running out of fresh data, where now we're mining old books for fresh data. If you buy into this whole world model thing, world models will be trained on world data. What happens when I drop a book, or when I drop something, or factory outputs, or machine outputs, that are constantly being updated and done.

And maybe the focus on trying to suck up everybody's data becomes less of an obsession than it is now. Now, I will tell you that China operates 10X the amount of factories, maybe more than we do. So back to the Chinese boogeyman and the yellow menace.

They have the ability, and they already do actually capture a lot of their factory data output. And they're using that to train models, and they'll probably use it for world models. We need to do a similar thing here.

This idea of just this clever little stunt of finding a loophole to go buy books in Europe and import them in here, some judge is going to close that up, and it'll be onto the next gimmick. Because there's a gimmick. I have this great book on alchemy, so I'm looking forward to when AI turns all the lead into gold and the whole gold market collapses.

It's going to be great. Yeah. Me too.

I like making notes in the margin and Just for me, this is heartbreaking, especially going to rare bookstores. Because when the copy is destroyed, and that copy is the last one in the world, the book and any margin notes are gone for good. And a providence loss, that money can't even remotely reverse.

And I do think that for me, that's an unconscionable thing. We live in such a society where rare doesn't deserve the honor to be treated as rare. I say this all the time, bubble wrap Stevie Wonder, because he's an awesome musician, right?

That's how we should be viewing rare books and talents and things like that, not shipping them off to an Amazon data center for the Dino Destroyer to destroy the book. Yeah, exactly. And the problem is that by allowing them to shred these books and convert them into digital and then feed them to the model, what they're doing is they're devaluing the intrinsic value of the content.

So they're treating it as a physical commodity, not as an intellectual commodity. And ultimately, that's the crying shame. And also, yes, please protect Stevie Wonder.

One more thing that I'll mention is the saddest thing in this 404 story is that they mentioned that the staff of the Amazon site in Las Vegas that was doing this destruction, they were internally wondering if they were going to lose their jobs soon because they were running out of books. And just think about that. And it still won't be enough.

Yeah. Another kind of fun thing I saw is they're only buying books with an ISBN because their database keys off the ISBN, and so there's a whole bunch of rare books that are not actually being encoded because they never were assigned this digital number. And again, that's just like, oh, God.

The whole thing is just a crying shame. Paul, I completely agree with you. It's not about the books, it's about the world.

Amazon started as a bookstore, and now it runs a warehouse where the last copy of a book becomes training data and then trash. Wow, I hadn't made that connection, but that- Yeah ... that makes me want to cry.

That's pretty ironic. Yeah. All right, guys, we got to hop to our next one.

Mike, what do we got for the third group today? Well, the next one is a post that Mitch wrote, and it's up on the Futurum Group site. But he's talking about, well, who's actually going to own these enterprise accounts, and who's going to be driving the conversation when it comes to AI?

Because, well, there's the traditional IT folks, there's been these kind of little tiger teams that people created, and then there's the business units that also fund these AI projects. And it's not necessarily a new conversation, but it sure is coming up in spades as we move into AI. So Mitch, explain.

Well, I've had untold number of conversations amongst vendors and also the buyers of who is the AI buyer in the organization. And while the CIO is not the ultimate buyer of everything technology anymore, we've gone way past that with bring your own device and lots of other reasons. Credit cards get you access to everything.

It still is a bit of a head-scratcher for vendors because it's also shifting at the same time. While purchases may have been made by the CIO to start with, they've also been being made in the business units, et cetera. So the data we have from our surveys shows roughly around 40%, just under, that live outside.

That's what the CIO and the CTO have. Everything else is outside, so you're talking about 62% majority of it, like a clear majority of it. And I have newer data that backs that up as well.

So a signal that's happening is also the hyperscalers have stayed out of the system integrator business, let's call it that, and let the Deloittes and people like that do that kind of work. But suddenly they're all investing or putting money or claiming they're putting money aside to build up those practices in their own organizations. So as part of their channel ecosystem, they've relied heavily, just like Microsoft doesn't have its own channel, go to market for the most part.

And saying, we're going to control not only the buying but the implementation of this. And I think a lot of that falls under when you are the implementer of a service for a customer, you're building these applications, models, agents, workflows, whatever it might be, that is stickiness. You hang around a long time.

So I think that's another angle. And by the way, you don't see Anthropic and OpenAI doing this, right? These are the big folks that have the dollars to put behind it.

All right. Well, Yvette, we've seen in the past that IT stuff rolls downhill, and usually it rolls downhill from some line of business somewhere, bought something, and then when they got tired of running it themselves, they called up the CIO and said, "Congratulations. " And then the CIO looks at it and goes, "Wow-" "Tag, you're it" ...

" So is that just going to play out here all over again? So let me just say, Mitch is spot on. Follow the money.

Since April, we have Microsoft, AWS, Google put about four and a quarter billion behind embedded implementation. They're like frontier, front teams, engineers. Microsoft put two and a half billion and 6,000 people inside customer teams, and then AWS turned around and did the same thing.

Google, again, same thing. So here's the thing. These same companies spent a decade avoiding all these low-margin services that their whole partner networks are still of, right?

With the Lloyd's and the PWC, all those, or Apple, right? And then turned around and funded the same bet inside 90 days. The reason why is because, as Mitch indicated, the buyer left IT.

They're not in IT anymore. And today's purchases, over two-thirds of them, I believe, sit outside of the CTO and CIO. So it's...

Oh, my gosh. I don't think people are thinking this all the way through because look, no one's thinking about the vendor that they hire, this engineer working inside your systems that they need access to, and scope to, and audit rights with approval gates that need to be set before they get there. So no one's talking about that matter.

And that's the crux of it. Because whoever proves a multi-model delivery first owns a layer, and nobody got that at a 6,000-person scale. I'll share with you a little lesson that I learned as a CIO because my folks would always be concerned about people buying their own stuff with credit cards.

I learned that there are three reasons people always go to IT. One is security. I want SSO for my whatever I'm building.

Two is I need data. I don't have access to the data that I need. That's kind of been curbed a little bit by MCP server access, but not necessarily fully.

And then, of course, the third thing is I need access to applications. Again, it might be solved somewhat by MCP, but not fully, or other agent kind of things. But still, I think those are three reasons that you're always going to go back to IT.

So it doesn't mean it's tossed back over the fence, tag, you're it, kind of the scenario we're saying, Mike. It could be. But I think here you're talking about not sort of whoever, Deloitte or somebody, owning this application that they built for you.

You're talking about Microsoft owning it. That's a different kind of relationship with Microsoft. And I think the CIOs are taking notice of, hmm, what does this mean?

Because they're working with them as well. So they've got a relationship, pretty strong one, too, with these vendors. And following the money, it's all where it gets spent, and that's meaning that's where the budget is for it.

Yeah. But there's the money and then there's what they do when they get inside your ish, right? So an engineer acting inside your company systems raises the same freaking control questions as when agents are deployed autonomously.

Access, scope, audit trail, the approval gates. A CIO who can't even govern his own patch management system, how do you think they're going to control that? That's my whole thing.

It's like we got folks who still get got through SharePoint vulnerabilities, and now you want to bring in these same companies that are selling you this stuff to run your stuff, and you can't even patch? Just come on, make it make sense. Make it make sense.

Move fast, break things- So let me make sense. Okay What we're really dealing with here, whether it's perception or reality, sometimes perception is reality, is that most non-IT departments and people want to go fast. They don't want the strings that come with dealing with the IT department.

We saw this happen with no code, low code. All of a sudden, I'm a sales guy, I'm a marketing person, I'm a business person. I need something fast and dirty, and it's going to take six months for me to do the requirements write up and get the approval and get the proctology exam.

And then it's approved, and by the time I get it, I don't even need it anymore. People want to just go. And they do.

And this is before AI. Obviously. Now you got these forward deployed engineers doing it, and everybody's got one it seems.

What does it say that you think me bringing in a third party is going to make me get it faster, better, cheaper than bringing in my own people? Well, to me, that points to a problem with the CIO and the IT organization. Heal thyself.

If I could bring strangers in that I got to pay extra for, because I don't think you could get it done quick enough for me or well enough for me, we got a problem, Houston. Yeah. There's a problem here, and that's really the underlying problem of this.

If I thought my CIO and IT organization would actually help me, I wouldn't bother with this other stuff. I'd go to them. It's the same thing with security.

Please don't call the security people, they're going to say no. I already know they're going to say no. We had to overcome that.

Yes, we can. Someone came up with a phrase around that once they became president. Yes, we can.

I don't understand. That's what CIO's got. Me too.

That's why CIO's got to- Yeah ... the job. Yes, we can.

Yes, we can. But there's a balance between yes we can and responsible IT management. Steven, you've been a long time IT guy, and at the end of the day, a lot of these projects are failing because nobody took into account IT fundamentals.

Absolutely. Absolutely. Fundamentals, security- Fundamentals ...

networking, storage, scalability, integration, all those things. But- Locking an S3 bucket. Locking an S3 bucket.

Locking an S3 bucket. Not making it open to the world. Hey, I'm feeling attacked here, okay?

Locking S3 buckets. The funny thing is, for all this talk of forward deployed engineers, inside IT, I think that as you're pointing out, it feels a little like foxes guarding the henhouse a little bit to have Microsoft come in and tell you how to use the cloud. But actually, the funny thing is, again, as a long time IT person, we saw this happen literally every technology that has ever been introduced to IT in the history of IT.

Personal computers, phones, everything, Windows. All this stuff was brought in outside the control of IT, against the will of IT, and eventually IT had to adapt. But it's not an IT thing.

If you look at other parts of industry generally. I was just reading a story about a car company, and they have a supplier that's coming in that's developing the tooling and developing the prototype parts and figuring out the manufacturing processes, and then they're going to ship all this stuff to the factory and install it and train the workers on it. How is that any different from a Palantir forward deployed AI engineer?

It's not. This is literally how every company does business. And yeah, it's probably bad.

Do you really want a supplier developing all the tooling and deciding which components go into your car? Hmm, I wonder which components they're going to pick. But it's the same thing.

And this happens in the steel industry, and it happens in the oil industry, and it happens in every industry. It's just good business to mess with your customers and make them dependent on you. Well, we have this in the model companies selling us token usage subscriptions at the same time.

There's no conflict of interest there. I think it's all about account ownership, and the big hyperscalers are competing with each other to say, "We all have similar capabilities. Yeah, we have different strengths.

" Whatever it is that you have to implement. You're not going to hire Microsoft to come implement a better calendar analyst skill. You're going to be building significant applications, agents, whatever it is.

So I think at the end of the day is these folks don't want to lose. They want to take their advantage. Google's going to bring their advantage of chip to software to model.

" And same thing to a degree, same degree with AWS and Bedrock and things like that. So I think that's really all about capturing the customer before they get too far down the road and suddenly they've lost because somebody else is in there first. We only have a minute, and I want to get Pawel in here because I counted no less than four grimaces.

So what were mine? Right. So a little bit against this kind of YOLO approach to buying stuff.

Because today, when you buy a hammer, so like a product, the only problem you can have, you hit your own finger, and you will be in pain, or someone else's head. But when you are taking AI tool, it's like inviting to your organization or home, someone from outside with no way of checking who you invite. In 99% of cases, it will be okay, but this 1% can be problematic.

You can have empty flat after that. So from this perspective, I would like to have some kind of security involved into that to just at least assess should I go with that or should I stop? And that's the question.

But we can't answer it because we're out of time. Yes. But we will continue this conversation I'm sure throughout the week.

As a matter of fact, we have this conversation every Monday to Friday at noon Eastern time here on Techstrong TV. You can catch the gang live. tv, Techstrong TV YouTube channel, Techstrong TV OTT channel for Roku, Amazon Fire Stick, Apple, iOS, Android.

If you got a screen, we're coming at you. You can check it out there. Gang, thank you so much for coming on here and sharing today.

It was a great discussion as always. Pawel, again, happy birthday. Thank you.

Yvonne, Steven, hopefully we'll see you next Monday as well. Mitch, keep writing, keep publishing, Mitch. We love it.

Mike, we'll talk to you later. Until then, everyone, this is Alan Shimel. We're out.