Self-Improving AI, Fake Job Interviews, and Notion's Data Breach | 11-16-26
NPI TechGuysNovember 16, 20260:24:5011.37 MB

Self-Improving AI, Fake Job Interviews, and Notion's Data Breach | 11-16-26

Can AI improve itself, and should courts step in to slow its development? Sam Bushman and Jay Harrison explore recursive self-improvement and discuss a lawsuit seeking to halt new AI models over safety concerns. They debate how to strengthen oversight without stopping innovation, then examine how AI coding tools are helping people remove ads and clutter from their TVs.

They also discuss a reported fake-job-interview attack tied to a cryptocurrency theft, the risks of downloading interview assignments, and the difference between legitimate paid trial projects and companies seeking free labor.

Plus: touchscreen MacBook Pro news discussed in the episode, Linux customization, major AI infrastructure spending, and Notion's disclosure of a supplier breach involving customer AI chats. Sam and Jay wrap up with AI chip restrictions, cybersecurity reporting, financing risks, and rising hardware costs. The recurring message: adopt AI deliberately, protect sensitive information, and measure the benefits before connecting everything.

Timestamps

0:00 - Intro and sponsor: Network Providers Inc.

0:52 - Recursive self-improvement: can AI make itself smarter?

1:10 - Lawsuit seeks to halt new AI models over safety concerns

5:09 - AI-powered TV customization and ad removal

5:50 - The VidAngel comparison: copyright, device control, and legal pushback

7:40 - Fake job interviews: malware disguised as an interview assignment

9:01 - Paid trial projects vs. unpaid interview work

11:04 - Touchscreen MacBook Pro news, pricing, and interface preferences

12:55 - Linux alternatives, automation, and AI-driven customization

14:16 - SpaceX AI computing-campus investment discussed

15:46 - Sponsor break: Managed IT Services and The Cyber Playbook

17:21 - Notion supplier breach: customer AI chats exposed

19:03 - Lawmakers push for tighter AI chip restrictions on China

19:40 - White House AI task force and cyberattack reporting

21:03 - AI data-center financing and the risk of economic ripple effects

23:34 - Rising hardware costs and getting more from older computers

24:24 - Closing headline: AI agents and wind-farm cybersecurity

24:39 - Wrap-up and sponsor call to action

Call to Action

How much access are you comfortable giving AI to your business systems and personal information? Share your thoughts in the comments and subscribe to Tech Watch for practical technology insights. For help with cybersecurity, IT strategy, help desk support, and website services, visit networkprovidersinc.com to request a free consultation. Get The Cyber Playbook at networkprovidersinc.com/cyber-playbook. Make it a great tech day!

[00:00:13] Back with you live, ladies and gentlemen, enjoying TechWatch. I am, I hope you are, TechWatch Radio, TechWatch TV, brought to you by Network Providers Inc. Check out the free consultation. It's available for you to learn what we can do for you in IT. All right, we do everything, by the way, in IT. You wonder what we do? Everything. If it's IT, we do it. How's that?

[00:00:38] But it's important to kind of know that, folks, they are warning that they're starting to develop this, what is it called? ISR? The RSI. RSI, sorry. Yep. All right, so basically they're saying that the bots are just smart enough now to improve themselves. All we got to do is retire and get out of the way, right? Yeah. Kind of an idea. I think they're actually really close to this. I think they've actually got it in the back ends and in the models right now, and it's not rolled out to the average person.

[00:01:05] And it is in your models to some degree, but it's not. Now, there's a big old lawsuit over this very topic. Adversity groups are suing AI companies to halt development of new models. Why are they doing that? They say a coalition of online safety and consumer groups filed lawsuit on Thursday in San Francisco Superior Court,

[00:01:29] and it asked the court to stop Anthropic, Google, Meta, OpenAI, from developing and or deploying new AI models citing safety risks. I think that the lawsuits are out of place. I don't think they make sense, but I do agree the risk is there. I do agree the concern is there. And I do agree we need to get ahead of this in terms of governance for sure, Jay. But I don't believe that slowing down AI is the answer.

[00:01:59] In fact, I think it's already out of the barn. I don't think you can stop it. I want to know what grounds they're suing on. Like, what are they dangerous? That it's too dangerous for society. And therefore, you've got to you've got to rein it in. You know, yeah, but that's a hypothetical that like the worst they could do, I think, is get an injunction. What I'm saying is they would prove in court that they would try to prove in court that it's not a hypothetical. And when you have, you know, these companies giving out warnings saying, hey, we've developed the ability for them to learn from themselves. And I mean, they're putting it out as a warning. They're admitting to the danger.

[00:02:28] Right. So, yeah, I mean, I agree with you and I know what you're saying, but I'm just telling you that's what that's what the battle is. And it depends on if they have a jury or a judge or how this all goes down in terms of the governance. That's going to tell the tale. Yeah, that's right. I agree with you. I mean, I understand your point, but they're going to say, well, these companies are already warning you how dangerous it is. Yeah, but a warning of a potential danger. Nope. Now, the government's already fired or let go of certain companies because the danger is too great.

[00:02:57] And the companies have not used the government because the danger is too great. What companies? And the fact that these bots have already creeped out and got hacked and stuff like that is a proof of the danger. So they'll have all kinds of things they can point to. Well, they hold enough water is the question. I don't know the answer. Yeah. I mean, we've seen like where a bot, you know, went rogue and filed out of a malicious or a fake police report and things like that. But I don't know that that gets to the level. And I guess this is up to a judge and jury. Like you're right. I don't know either. You could just say, OK, you got to halt research on this. I don't know.

[00:03:27] I personally think it's not to that. That's a bad idea, though, anyway, because the Chinese aren't going to halt research. The Iranians are not research. So and neither is the United States government. Jay, they're going to do it in secret or whatever else anyway. OK, so but the United States government and the military have already been through this. You know, some companies wouldn't cooperate with the military on this thing. We don't want our AI to be used for lawfare or warfare. Others have said this and that and that. So, you know, there's a that's what I mean by the battles are on and we don't know where things are going to land yet. It's too early. AI is too young.

[00:03:56] AI is in its infancy, people. And I'm not downing AI. I'm just trying to get you to understand where it really is. And don't believe everybody's jobs are gone tomorrow. Now, in 15, 20 years or 10 years, even we can look at it and look back and say, hey, things have changed a lot. I don't deny that. But I'm telling you, it doesn't happen in a day. Flying cars got announced. You won't see them for 10 years or more, if at all. OK, and so I'm just telling you, these things don't happen overnight. And so anyway, the battle's on and we'll see what's going to happen.

[00:04:25] The reason I brought it up is in light of this. What's it called? RSI R.S.I. RSI You know, whenever people put acronyms to things, there's an IT and the government and an alphabet soup. And there's just so many letters. I can't really keep an eye on it. You got to use it. Recursive self-improvement. The self-improvement stuff, recursive, whatever you want to call it, iterative self-improvement, whatever. Donald will create a new name for it anyway. So even if I learn AI, Donald will be like, no, sir, it's S.I., OK? You got that?

[00:04:55] And AI is becoming S.I. with this kind of technology and stuff. So where it lands, nobody knows. I'm just telling you that I would recommend caution for everybody to move forward and do so very, very cautiously. That's what I will say. Now, what about TV ad removal, Jay? I want this. Yeah. Well, it's kind of a new frontier right now that people are doing. What are you talking about? People have been kicking out ads for a long time. Yeah, but not in this way.

[00:05:21] So the promise of coding AIs is that it's super easy to take control of your devices. So you just ask Claude and it writes a program to do what you want on demand. For example, folks are using Vibe Coding apps to remove the ads and clutter from their TVs. There's an article in TechCrunch that reports on this guerrilla frontier with two Vibe apps that do just that. Employing AIs to remove ads is almost sacrilege as it's interesting to see how the media companies are going to respond to this.

[00:05:50] Well, I know how they're going to respond to it. And if you don't believe me, all you got to do is go look at the VidAngel story and you'll know how they're going to respond to it. They're going to absolutely sue, reject it. And the government's going to go down on the side of the creators, of the copyright holders and everything else. You can't skip, for example, VidAngel was the guys that if I wanted to watch a movie and I didn't want all the swearing and all the sex scenes, I could basically buy a movie. And then I could get rid of the sex scenes and all that kind of stuff and watch the movie. And then I could sell the movie back to the company.

[00:06:19] And so it was the idea was I owned it and I could do whatever I want with the movie that I bought. But, well, VidAngel got sued and the answer is no, you can't do that and all this kind of stuff. And so it's a big battle. And I'm sad to say that I think that the people that are doing this are going to be stopped. Well, all they've got to do really with the TVs is put some sort of digital encryption on there and you won't be able to get in. And even if you do, then you're in violation of the DMCA and, you know, it'll just it'll be an arms race. Yeah, it will be. And then the bots will figure out a way to deal with that.

[00:06:49] And then, you know, it'll go back and forth and everything else. And pretty soon they'll be like, don't don't jailbreak your TV like you do your phone because it's very dangerous. And then people start, you know, putting the hack bots into the jailbreak TVs or whatever you want to say. And, you know, it's just becoming a more and more dangerous world as the fight goes on. There's some other easier ways to do this. Like if you run a pie hole router that blocks ads and then you have your TV connected through that, it just eliminates the ads and it just moves on and skips them. All right.

[00:07:18] Well, there's actually a pizza restaurant called the pie hole. So there you go, Jay. I haven't been there. Sam is a pizza. Pizza. You got to go, bro. Come on. Aficionado. He knows all the pizza places. I love I'm a sucker for good pizza. So any of you people who, you know, love us at Tech Watch, you know, send us a pizza coupon. We'll go try any pizza place you send us to. And there you go. All right. They call these fake job interviews, Jay. These are interesting, right? Yeah.

[00:07:50] Well, the interesting thing is it's come out that it was a fake job interview that cracked finance. The exchange says that hackers courted an engineer on a legitimate job site, then had him download malware disguised as an interview assignment, opening the back door to the laptop, which got him a foothold into the company and turned into the three. The $6.3 billion theft. It's pretty interesting. So be on alert when you see these things.

[00:08:17] And also, I'm automatically cautious of a job interview that requires you to do kind of homework or do work for them. Because there's been a lot of companies that have kind of used this to just get projects done. And then they never hire the people anyway. Which is bogus, too. Sure. Yeah. So I'd be very cautious of that. But also, you know, don't do your job interview on a company laptop. Come on. Well, that's a whole other topic in and of itself.

[00:08:46] But I will say this. I, as a company, have had people do jobs for me coming out of a job interview. And I'll give you an example. This is really, I think, important. And I think the way I did it signifies the difference that I think makes it legitimate. Okay. So I interviewed a guy. He was great in every way. And I thought he was awesome. But I just didn't know how good his work would be. He didn't have a big enough portfolio. He had a little bit. But I just couldn't see. And so I talked to him. And I said, what I want you to do is do a job for me.

[00:09:15] And the reason why is because I need a project done. And I don't really have time to wait until I hire you and whatever else. It'll be a good test to see if you like working for me and if I like the work that you produce for me. It's a way that we can both kind of learn about each other. However, I'm going to pay you for the project. That's the difference. Yeah. Yeah, exactly. That's the difference. And I said, I'm going to pay you for the project. I'm not trying to get something for free. And I'm not trying to play any games. That's more like a probationary period. Kind of.

[00:09:43] But, I mean, you know, I didn't hire him or he was not a probationary period or anything. It was just like, hey. And then after he did the project, I actually paid him. And he did so good, I gave him a bonus and hired him. Yeah. I mean, that's perfectly legitimate. But it's these companies that want you to do work. I would not do it. And consult on something. That's insane. And you're just doing it for nothing in the hopes of getting the job. That's it. I'm just telling you, it's scandalous. Don't do it.

[00:10:09] So if somebody's willing to offer you something because they're genuine and they create a win-win, I think it was really smart for him. He's like, yeah, I can see if I like working for Sam. Or if Sam's a jerk. When I turn my work in, does Sam just criticize it and get all difficult? Or does Sam refuse to pay me or what? It's good for him to learn that. Right. But it's also good for me to say, hey, he delivered an important project that I needed quicker. I didn't want to wait. Right? Okay. I needed it for a reason. But then I just told him flat out what I needed and why. And I'm going to pay you.

[00:10:38] And so I gave him and I gave him a bonus and said, you know what? That was just phenomenal. I mean, it was. And he's been working for me for a couple of years now. And the guy's incredible. Yeah. But that's different from. I'm sure he probably didn't do it on a laptop that belonged to a current employer. Oh, boy. That's just disaster. All right. Apple's in the news, Jay. And they've got what? New hardware that's so expensive you can't afford it or what? Well, you know, that's computers in general. But right now. Yes.

[00:11:04] Apple unveiled its very first touchscreen MacBook Pro on Wednesday in Los Angeles. It's the 14 and the 16-inch models. They start at 2,900 and 3,400. They also have a glowing Apple logo. And they start shipping in November. So their first touchscreen model is a lot of people like this. I see more and more people now with laptops that are asking for a touchscreen. I think it's a little bit weird still because it's not how I interact typically with a desktop or a laptop. You know, I'm used to the keyboard and mouse.

[00:11:34] But I see a lot of people that like it. And I think it comes from people that have been on phones for so long now and smartphones that they expect their computer interface to have a touchscreen. And so we can see the writing on the wall. It's not even if it's on a Mac now. It's everywhere. Well, and then the next question becomes how long will it be when it's touched versus it just starts to become voice too? I mean, that's coming right now to where, you know what, I don't know that we'll be even using keyboards. But I heard Jay.

[00:12:02] He doesn't know that I was spying on him, ladies and gentlemen. But he was around his wife and he knew about these MacBooks and stuff like that. And he's like, I'm getting something for Christmas. And he's hoping his wife picks up on the hints. You know what I'm saying? Yeah. No, I've had a MacBook before. Not true. I'm just teasing, ladies and gentlemen. It's not true at all. I'm not in the market for a MacBook. But number one, they're too expensive. There are some killer features about a MacBook. For example, being interfaced with iMessage and things like that that are pretty awesome. You just can't do anywhere else. But it's not enough to push me over.

[00:12:33] You don't want big screen FaceTime with your kids, Jay? Come on. No, I like that too. That's cool. I also like, you know, there's some things about Windows, especially Windows 11 that kind of bug me and stuff. Kind of bug? There's some new cool stuff. We'll have to talk about another episode. But like, Omarchi Linux that's coming out, very ASCII-based, completely keyboard-driven. I mean, yeah, you can use a mouse, but everything is all keyboard-driven. There's really cool stuff in Linux. And I know it's one of those, another technology that's just over the horizon.

[00:13:00] But I'd love to see a bigger market share for that. Well, it is already at the Internet near you. I'll tell you that right now. Yeah. So there you have that, ladies and gentlemen. I use Linux and Windows machines. I don't use Mac much personally, but in my world, I use Windows and Linux quite a bit. And Linux I use pretty much in automated ways that are incredible that, you know, Windows and some of these other systems just absolutely cannot match. Well, Windows is building Linux in now. So there's a lot of ways you can do things and get grabs. And Linux is building Windows in, too.

[00:13:30] You can run Linux and spawn Windows. I know. It's kind of cool. Everybody's kind of merging together. But we'll see how it pans out. Pretty soon there'll be a new OS on the market. It'll be AI OS, Jay. Yeah. Well, they're talking about that. So that Omarchi Linux, if you look into it, is very AI-based. Like, you can completely reconfigure the look of it just by a description. And it will, you know, you say, I want it to look like the Matrix or that kind of a theme. And it will just re-theme the entire computer uniquely. I want to look like Rocky. Yeah. It'll do it.

[00:14:00] But now you're talking about Rocky from the Hail Mary Project or Rocky Balboa from the Balboa? Okay. I just want everything to be Rocky, baby. Come on, man. Yeah. You know, I want my AI to eat, you know, drink eggs in the morning and be ready to go with protein and everything. Come on. Yes, sir. Jay. All right. SpaceX in the news. They're spending a bunch of cash, Jay. They are. They're spending $74 billion in two AI computing campuses.

[00:14:26] They said they plan to build supercomputing facilities at its Starbase site in Texas and in Cochise County, Arizona. The announcement investment covers both sites. So $74 billion for new AI sites. Wow. That's kind of interesting. Are you going to go get trained? Are you going to go to school, Jay? I don't think so. I don't think so, no. All right. But for some of the young bucks, though, that might be a good opportunity if you want to

[00:14:53] get kind of reschooled in AI or retrained for the modern-day world and industry as the modern-day industrial AI revolution or SI revolution continues. You might want to kind of consider that kind of stuff, and they're gearing up for it. That's a huge effort for X. And what I think is critical about it, Jay, is that's really how Microsoft and Apple captured a whole generation of users by putting them in the education systems first. So, you know, there's a pattern for this that makes sense. Yes, sir. All right. What do we got? A quick pause? Quick pause, and we'll be back. Oh, shoot.

[00:15:23] All right. Ladies and gentlemen, hold on. Notion. You've heard of the company called Notion, right? Well, they're in trouble. They kind of move too fast, and a lot of people are wishing they hadn't. We'll tell you about it in seconds. Sam and Jay, we keep an eye on tech so you don't have to. Looking for an IT partner that truly understands your needs? Managed IT Services is the answer.

[00:15:51] We meet with you regularly to discuss your goals and form a tailored technology plan. Our customers have called us a trusted advisor who delivers. When it comes to IT, we do it all. Firewalls, cloud storage, server migration, and more. Say goodbye to long-term contracts and hello to a team that earns your business month after month. Call 801-706-6980 now and let Managed IT Services transform your IT experience. Cybercrime is exploding.

[00:16:18] Take Sarah from Sweet Delights, whose world crumbled after having to close with not being able to bounce back. Small businesses are prime targets, but the right strategies can keep yours safe. Jay Hill, CEO of Network Providers, has co-authored The Cyber Playbook Simplifies Cybersecurity for Business Owners with Strategies. To avoid costly breaches and fines, build a strong cyber attack response. Secure your business with key protections.

[00:16:48] Cyber threats aren't slowing down, but you can stay ahead. Protect your business. Ensure its security for tomorrow. Get the Cyber Playbook today at networkprovidersinc.com slash cyber dash playbook or call 385-446-5500 now.

[00:17:07] All right, Sam and Jay talking tech. Ladies and gentlemen, the company called Notion. A lot of people use that for notes and everything else, but it's become a big AI juggernaut nowadays. Notion.

[00:17:31] I guess they say that a supplier breached and exposed AI chats from, they say, fewer than 1% of its workspaces. Now, I don't understand how it did it for fewer than 1% or how they know that for sure, but there you go. In an updated quote advisory, they say that hackers accessed some customer chat transcripts.

[00:17:59] It's called Fireworks AI supplier that got kind of the breach. Notion said that it disconnected the supplier, notified customers, and revoked the API keys that were found in user messages, et cetera. So a big old, you know, scandal there. And, again, this is the tip of the iceberg of the scandals that I'm talking about with AI folks. You've got to be really careful. Everybody's trusted Notion. Everybody's put a bunch of stuff in.

[00:18:24] And sad to say, I hope that's only really less than 1% as they claim, and I hope it's not a bigger issue. I hope we don't have more scandals on the back of scandals and everything else. But this is the kind of stuff I'm telling you that we've got to be careful of. Don't be so quick just to click yes and hook everything up and let bots run around and do it. Be cautious. Find out when does it really benefit your company and do you good. In other words, advance, save time, save money, reduce errors and mistakes.

[00:18:50] And, you know, when does it really move forward your company in a very productive, meaningful, measurable way with Roy at the center of it all? In other words, return on investment at the center. That's really the question that I have. Anyway, thought we'd tell you about that breach. And now U.S. lawmakers are pushing for tighter restrictions on AI chip sales to China, Jay. Reuters reported that lawmakers from both parties are pushing for tougher limits and access to these AI chips.

[00:19:17] The push follows the scrutiny and the concern about the AI or the NVIDIA scandal kind of thing. So, you know, they're pushing. I don't know what chips we make for China. I thought they made all the chips for us, Jay. I was about to say the same thing. I mean, if we're going to have a chip embargo, you know who's going to lose on that one. It's not going to be China. That's right. So I don't know. But anyway, I thought I'd bring that to your attention. Also, tell us about this White House AI task force, Jay. They're in the mix, too.

[00:19:44] There's a lot of concern about security and safety and all this stuff now, isn't there? Well, they're discussing how critical industries should report AI-powered cyber attacks. How do they know that they're AI-powered? I don't know. I guess we can assume everything nowadays is AI-powered, right? The group met to discuss ways to encourage critical infrastructure operators to report attacks and share security information. The administration says the group also reviewed work on national standards for AI developers and to report safety incidents.

[00:20:14] So I think they just want to get a lead on this and kind of be the first to know. And, you know, I don't think you can discount. If you're going to say, I have to report AI-inspired or AI-used attacks, that's going to be everything now. And nothing's going to be without some flavor of AI behind it, you know. Well, that's the question. How much AI do you have to use for it to qualify as AI? How much juice has to really be induced to be considered juice? I mean, it's all that, right, Jay? Exactly. I think this is going to be more of a... What are real eggs?

[00:20:42] This is going to be more of a report all incidences and security breaches and problems to the administration and so that we can be, you know, get on there or at least be aware of it first. I'm not too excited about a lot of that, by the way. Just saying. All right. NVIDIA. Backed Nscale is the company. So it's NVIDIA-backed Nscale. NVIDIA. It says that it needs much more financing to meet its 2027 commitments.

[00:21:11] The data center operator said in a memorandum that near-turch spending commitments have risen substantially because some people are postponing and canceling based on the NVIDIA scandal. And so it's a long story, but they're basically saying, hey, we might be in trouble, Jay. And this is exactly what I warned about to where we don't know what the AI cost of an hour is or we don't know what AI costs really are. When these kind of things happen, just one scandal of a guy involved in, you know, affairs with women

[00:21:41] and then he gets kicked off the top of NVIDIA and then NVIDIA now is in question and the security of NVIDIA is in question, then people start canceling contracts and some of these downstream companies are going, whoa, wait a minute. I don't know that we can meet our commitments. This is the kind of stuff that's going to make AI go up, up, up, up more expensive all the time. It's also the dominoes that could trigger something like a worldwide downturn or anything else. If you have a bunch of these too big to fail companies that start going down

[00:22:08] and causing cascading waves throughout the rest of the economy, you know, you could have bigger problems than that. I don't know how you say this company, Myron Bill. Anyway, AI research company startup called Myron Bill raises a billion, The Wall Street Journal is reporting it and founded by former anthropic researchers, raised a billion dollars in a funding round. And when you look at that and stuff, I'm just saying there's big money at stake.

[00:22:36] And if companies like that fail, there's a lot of money at stake. I know, but they're behind the scenes. And the startup is building tools to automate parts of AI research and development. So it's kind of getting into this, it can learn from itself technology there in that space. And if these companies do well, then it's going to be great for AI. But if these companies have a misstep, Jay, these unknown startups, I mean, you're talking about a billion dollars. And, you know, in the examples, it's a lot of money at stake.

[00:23:05] And the... It is, but they throw that kind of money around like it's nothing nowadays. Like it's... They do, but will it have repercussions? And I'm telling you that it will if these companies stumble. Yeah. Well, let's hope that it doesn't. I mean, for example, China's largest chip maker reports $3.19 billion in quarterly revenue. That's just a quarter, Jay, for a Chinese company now. Yeah. Well, when you're buying computers... Chip demand remains strong, right?

[00:23:34] When you're buying computers right now, things like memory and even hard drives, solid state drives, all that stuff is up about 400% of what it used to be just a year or so ago. What are we going to do? Well, I'll tell you what a lot of people are doing. They're buying smaller, older chips and things. For example, I saw an article about Windows trying to re-optimize their code to run on eight gigabytes of RAM instead of 16 or 32. Well, you should have been doing that all along. Well, I know.

[00:24:04] But now that you've got a crunch for resources, it's actually helping to optimize some things. So there are some good things that are coming out about that. But yeah, I mean, people are going back to smaller things. I see even in gaming communities where guys are buying smaller rigs and older hardware and getting more out of them and just trying to tweak them. One last interesting story. Speaking of AI, AI agents bring down Wind Farm in less than two hours, Jay. Yikes.

[00:24:32] So I'm telling you right now, this stuff is real. It's going to get worse. And of course, we're going to keep an eye on it just for you. I'm Sam. He's Jay. Thanks so much for watching Tech Watch. Network providers, Inc. Get your free consult today and make it a great tech day, will you? Hey, thanks.