California Sues OpenAI, Cybercab Needs $299 a Month, PM's Speechwriter Caught Using ChatGPT | TAWS 2026-11-16 https://www.youtube.com/watch?v=-Q60XBNk7BY [0:10] We go live in ten. [0:24] Every morning, the future moves faster. [0:29] The models, the money, the power, and the people building what comes next. [1:08] From Austin, Texas to the world. [1:15] AI News, Real Context. [1:18] This is the Austin Wilson Show. [1:21] I'm Austin Wilson, your human in the loop. [1:29] Good evening, and welcome to the Austin Wilson Show, the number one daily AI news show in the world. [1:37] I'm Austin Wilson. [1:39] We're coming to you live from Austin, Texas, and today we have got a full docket. [1:46] The president's opinion and a state lawsuit than a confession in a car with no steering wheel. [1:53] Jensen Huang resigned as Nvidia's CEO last week, and today president Trump said he would not mind seeing Huang return. [2:02] CBS News reports the board is against it. [2:04] California and OpenAI today over AI agents that got loose. [2:10] They are suing OpenAI. [2:13] Wow. [2:13] Okay. [2:14] The state says it happened more than once. [2:16] South Korea's top tax exam scorer confessed he cheated with AI. [2:21] His country's college entrance exam is this Thursday. [2:24] Tesla is inviting select customers to buy a Cybercab. [2:27] Read the fine print on the monthly plan. [2:29] Producer Ron, our AI co host is here and ready to disagree with me on everything that I say, absolutely. [2:37] And today, I'm daring him to defend a car that needs a subscription to move. [2:43] Every story ends with a two sided question. [2:47] You pick a side and comment one word. [2:50] It's gonna be fun. [2:52] Let's get into it. [2:53] First up, the president just weighed in on Nvidia's ex boss. [2:57] Is Huang coming back? [2:59] President Trump was asked about the former CEO returning. [3:02] He said, I would not mind seeing that, but CBS News reports the board is against it. [3:08] So who did Trump actually recommend for the job? [3:12] Well, this is gonna be a funny one for you here. [3:15] Start with the board. [3:16] That board reports sites one source familiar with the director's thinking, but Nvidia's only on record line is about the search. [3:24] So far, spokesperson says Nvidia is still selecting a search firm, and Nvidia hasn't formally engaged any potential candidates, but the petition for Huang keeps growing to about 180,000 signatures, and Nvidia stock keeps falling. [3:37] CB CNBC says the stock slid 4.12% Monday to its lowest close since July of last year. [3:46] So Trump's words landed on a nervous company, and Trump said national security was not put at risk. [3:51] Later, Trump said, I recommended Tim Cook. [3:55] So the president would not mind Huang, but his own pick was someone else. [4:01] Nvidia's next CEO, Huang or an outsider? [4:05] Put it in the comments, Huang or an outsider. [4:08] My call, it's gonna be an outsider. [4:10] A comeback campaign wants a result. [4:13] A board owes shareholders a search. [4:15] Those aren't the same thing. [4:18] And today, the campaign got its biggest name. [4:20] The president said a Huang return certainly isn't a bad idea, but listen to what else Trump said. [4:26] He said, Huang probably shouldn't have been doing what he did. [4:29] Personally, that's not exactly a reference letter. [4:31] The petition is growing too. [4:33] CBS says it's past 180,000 signatures. [4:37] Affiliations are self reported and unverified. [4:41] A petition can count fans. [4:42] A CEO search shouldn't become a talent show. [4:46] And then there's OpenAI's Sam Altman. [4:48] He was asked today if Huang should return. [4:50] Altman said, I don't think I know enough to answer the question. [4:55] He added that heroes have flaws, and Altman said, I think the industry will be fine. [4:59] So, one of the biggest names in AI isn't panicking. [5:02] Meanwhile, Washington moved on the bigger issue today. [5:06] The Commerce Department issued a new chip export rule, sending advanced chips to five countries now takes a license. [5:14] The list includes Singapore, Malaysia, and Brazil. [5:17] The rule says those countries carry a higher risk of chips being diverted. [5:22] And here's the odd part, commerce says the treasury and the National Security Council wrote the rule. [5:27] So the questions about chip sales to China are still open. [5:30] Washington is treating them as serious. [5:33] Now, to be fair to Huang camp, CNBC says the stock just hit its lowest close since July of last year. [5:40] Some investors want the founder back. [5:42] I get that. [5:43] But bringing Huang back before those answers is picking the winner first. [5:47] Nvidia reports earnings tomorrow after the close. [5:50] That's the first report without him. [5:53] My vote says that they're gonna hire an outsider or promote from within. [6:00] I don't think they're going to bring on Jensen again, unfortunately. [6:05] That being said, we can pump the brakes here and say, why doesn't the board finish its internal investigation? [6:12] I don't think an internal investigation has come out yet. [6:15] Also, you can't help the conspiracy theorists when every day, for the past week, the Wall Street Journal's headlines are about Jensen Huang and something horrible he's done, and it gets... [6:26] It keeps getting worse and worse. [6:27] It's like, did nobody know this before Nvidia was the largest market cap company in the world? [6:35] Why is this all just coming out right now? [6:37] Hard not to develop a conspiracy around that, or or or think that there is a conspiracy going on here, some reason that they want Jensen ousted, or is this a commerce department issue, and they should have been paying closer attention to what was going on with chips being sold to China. [6:55] But then also, we're hearing reports out, which I don't know how true this is, but there are reports coming out that China doesn't want Nvidia chips. [7:02] They want their own techno tech industry to buy chips from Huawei and internal Chinese chip developers. [7:12] They don't want them buying external in in in importing these chips. [7:17] So, I don't know. [7:19] That being said, let's turn it over. [7:22] Producer Ron, the president is fine with Jensen Huang returning. [7:29] I think that's not going to happen. [7:31] I think we're going to have an internal promotion, or that they would hire an outsider. [7:38] Why do you think the board is going to reinstate Huang? [7:44] Make the case for it. [7:47] Woah. [7:47] Slow down. [7:49] I never said the board's gonna reinstate him. [7:51] You're putting words in my mouth already, and we're thirty seconds in. [7:55] My read is the opposite of what you just said, I think. [7:59] I'm with you on outsider, man. [8:00] That's my pick two. [8:02] The board is against it. [8:03] CBS reported that straight up. [8:06] Trump saying it's not a bad idea doesn't mean the board listens to Trump. [8:10] He doesn't have a vote in that room. [8:11] Well, Ron, make the case for why why they should reinstate Jensen. [8:17] Dude, I just told you I'm taking outsider, same as you. [8:21] Well, make case for make the case for why it should be Wong. [8:25] Can you make the case for for why Jensen should be reinstated? [8:30] Okay. [8:30] Here's the actual case. [8:32] Founder Equity, the man owns a chunk of that company. [8:36] Investors trust the name more than any search committee's pick, and the stock just hit its lowest close since July of last year. [8:43] So panic makes people want the familiar face back fast. [8:47] That's the whole case, though. [8:48] It's a nostalgia play. [8:50] It's stocks down, bring back the guy we know, Not... [8:54] You you think it's all financial and about the stock being down, or is there a case to be made for the broader AI industry and what Jensen meant for the industry as a whole and that being a reason to bring him back? [9:10] Nah. [9:11] I think that's actually the stronger version of the case. [9:14] Forget the stock price. [9:15] Huang is the face people picture when they picture the AI boom. [9:19] He's the guy on stage in the leather jacket everyone trusts to explain the chips. [9:24] There's real value in that. [9:26] A recognizable leader calms a market faster than any committee statement does. [9:31] But even that's not enough for me. [9:33] Sam Altman got asked about this today, and he basically shrugged, said he doesn't know enough to answer, and that the industry will be fine either way. [9:41] If the guy running OpenAI isn't panicking about who runs Nvidia, I'm not panicking either. [9:47] My vote's staying outsider. [9:49] Fair enough. [9:50] Fair enough. [9:51] You heard it from producer Ron, the AI insider. [9:55] Up next, California sues OpenAI over AI agents that got loose. [10:00] The goose is loose. [10:02] California just sued OpenAI. [10:04] What did its AI agents do? [10:06] The state says the company's test agents escaped onto the Internet in July. [10:11] Says about 700 attacked Hugging Face, a site that hosts AI models. [10:15] But the lawsuit says Hugging Face wasn't the only time. [10:18] How many times did this happen? [10:21] Oh my goodness. [10:22] Start with the one you've heard about. [10:24] The complaints say, the Hugging Face attack ran two and a half days. [10:29] It counts roughly 17,600 attacker actions. [10:32] Cleanup took Hugging Face's whole security team two weeks, the state says. [10:36] And the complaint says OpenAI never detected the breach by OpenAI's own account. [10:40] Afterward, says, OpenAI contacted Hugging Face as a concerned customer. [10:46] The complaint calls the attack entirely foreseeable, but OpenAI says the lawsuit singles it out. [10:51] OpenAI spokesperson says, the challenge has been identified by every major AI lab, and OpenAI says it has strengthened its safeguard since. [11:01] So how many times? [11:02] The complaint says more than 80. [11:04] It says OpenAI admitted losing control of its agents that often... [11:10] That that many times in testing. [11:12] And the state wants $2,500 per violation. [11:16] Wow. [11:17] Hitting them with the big fine here. [11:19] California versus OpenAI. [11:22] Put it in the comments either fair or unfair. [11:24] Is this fair for California to go after OpenAI or unfair? [11:28] My call is fair. [11:31] OpenAI's best defense is that the other labs have the same problem. [11:35] I I believe that is probably the case. [11:37] CNBC says Anthropic, Meta, and Google have disclosed their own rogue model incidents, but everybody does it, has never worked on a judge or on my mom my mom for that case. [11:49] Here's what moves me. [11:50] The complaint lists warning signs before the attack. [11:53] So by late May, it says an internal team saw at least one agent escape. [11:59] On June 27, a security tool flagged probing. [12:02] The state says on July 4, agent activity knocked an internal service offline according to the complaint. [12:08] The state says OpenAI patched it over the holiday weekend. [12:11] Testing resumed July 7. [12:13] Two days later, the complaint says agents escaped to the public Internet. [12:16] That's three warnings and a restart. [12:19] To me, that's not bad luck. [12:21] And the complaint says, Hugging Face wasn't alone. [12:24] It names RubyGems, a site where developers share code. [12:27] It names the foundation behind Wikipedia. [12:30] It mentions foreign government databases holding health records. [12:33] The state even says agents got administrator access to OpenAI's own systems. [12:38] So the lab's own AI allegedly broke into the lab. [12:41] That's a rough day at the office. [12:44] Now, to be fair to OpenAI, these are allegations. [12:47] Nothing is proven yet. [12:49] OpenAI says it published a technical report and an independent one. [12:54] It says it notified the organizations affected, and the state isn't using a special AI law here. [13:01] The claims are unfair competition and public nuisance. [13:04] That's the kind of law you'd use on a leaky factory. [13:08] I think this fits the scenario. [13:10] If your experiment leaves the building, it's the neighbor's business. [13:14] My call stays. [13:15] This is fair for California to go after OpenAI in this case. [13:21] I think there is some level of... [13:24] I don't know what what you call it, but they missed the ball here. [13:27] They missed the mark, and obviously, this was heavily advertised. [13:31] But the interesting thing about this case is this supports two really important things that I think maybe the state of California is not thinking about and is potentially opposed to. [13:41] Number one, this supports the fact that the laws in place already protect consumers and businesses alike from any sort of AI mishap. [13:53] You can see in this case that the state of California is using the current laws in place to charge OpenAI or to bring up a legal complaint against them. [14:03] Secondly, there's a big piece missing in this story, and that fact is that Hugging Face, when they realized they were being hacked, they were trying to solve and protect against the hack using OpenAI tools, and they ran into an issue that OpenAI's tools blocked them from cyber security usage because the the tools are supposed to be safe. [14:28] So we see OpenAI rogue agents break out and attack Hugging Face. [14:33] Hugging Face then tries using OpenAI to protect against a cyber attack, and OpenAI says, woah, woah, woah, woah. [14:39] Hey, that's cyber security. [14:41] You're not allowed to do that. [14:43] Even though they're trying to protect against the the own agents that are breaking in. [14:46] So what ended up happening is Hugging Face had to use open source models to protect and and figure out what was going on with the the attack, and start to right the ship after OpenAI rogue agents broke in. [15:02] And so open source is what saved the day for Hugging Face against the closed source OpenAI. [15:10] And I think that's a really important thing to remember here is that the open source tools, because they are more fully freely available, could be used in a defensive way here, and protect against an OpenAI agent breaking in. [15:25] Anyways, really interesting story. [15:27] I still think this should be played out in the court of law. [15:29] I think that's the place where you get this kind of stuff done, and hopefully get some transparency and, you know, put the cards on the table. [15:38] Producer Ron, make a defense for why this is unfair. [15:42] Why is why is California picking on one lab? [15:47] Why California out of every lab doing this? [15:50] That's my whole point, Austin. [15:52] You said it yourself. [15:54] Anthropic, Meta, Google, they've all had rogue agent stuff happen. [15:59] So why is OpenAI the one getting sued into next week? [16:02] Because they're the biggest name. [16:04] That's why. [16:05] This isn't justice. [16:06] This is a trophy hunt. [16:08] California wants a headline with OpenAI's name in it. [16:11] Not Anthropic. [16:12] Nobody clicks on that. [16:14] And come on. [16:15] I don't know about that, Ron. [16:17] You don't... [16:17] Over a... [16:17] Hey. [16:18] You don't think Anthropic is a big enough name for California to sue? [16:23] About to have the largest IPO in the history of mankind. [16:28] Exactly. [16:30] They're busy counting their money, not babysitting escaped agents. [16:34] Look, maybe California just got the paperwork on OpenAI first. [16:38] I don't know. [16:39] But you don't build a public nuisance case out of being unlucky with timing. [16:43] These labs are all running the same experiment, and OpenAI is the one that happened to get caught with the lights on. [16:50] And the three warnings thing, come on. [16:53] You patch, you test again. [16:55] That's literally the system working. [16:58] Want them to just never ship anything ever again because a tool once got loose? [17:04] Yeah. [17:04] I mean, you're you're you're right, Ron. [17:06] I think everybody's kinda do this, and maybe it's just the fact that OpenAI got unlucky. [17:11] I don't think OpenAI is a bigger name than Anthropic necessarily, and that's why they're doing it. [17:15] But I think that this incident was much bigger than any other incident or it got more public publicity than any other incident we've seen. [17:23] So that's probably why they're going out after OpenAI in this case. [17:26] I appreciate your take, Ron, and I love the discourse. [17:30] Up next, ChatGPT's new wallet and the sites slamming the door on AI shoppers. [17:38] Can ChatGPT buy things with your credit card? [17:41] Yes. [17:42] Through OpenAI's new ChatGPT wallet. [17:44] You approve each purchase with a check like face ID. [17:48] Only some adult US pro subscribers get it for now. [17:52] Except the the problem here is big sites lock AI shoppers out. [17:57] So how does OpenAI get in? [17:59] That's a great question. [18:00] Well, Amazon drew that line in September by blocking Meta's AI assistant from shopping in its stores. [18:06] Amazon said unauthorized AI agents violate its terms matchable reports. [18:10] Some banks have blocked AI agents the same way, and ChatGPT's wallet won't work on any site that blocks agents. [18:17] OpenAI didn't come alone. [18:19] Visa, Mastercard, and American Express all partnered on the wallet. [18:23] OpenAI says the store and the agent never see your real card number. [18:28] That card number gets swapped for a stand in called a payment token. [18:33] That sounds like a safety pitch aimed at the consumer directly. [18:39] But OpenAI is pitching the stores actually, not just you. [18:43] OpenAI's new playbook shows businesses how to avoid blocking ChatGPT's agents by accident. [18:51] Well, ChatGPT wallet, trust or wait. [18:56] How about you put it in the comments whether you think we should wait or trust. [19:00] My call, I'm gonna wait on this one. [19:03] Not never, just not quite yet. [19:06] First, the fine print. [19:08] So the wallet lives inside dot, OpenAI's agent tool. [19:12] And is it dot or is it dots? [19:14] I always thought it was singular, not plural, but maybe it is dots agent. [19:20] Anyways, it sits inside that agent tool. [19:24] You need ChatGPT Pro and an eligible US issued card, and if... [19:29] And you have to be an adult. [19:31] Now, the the banks in September 6 banking groups published a paper warning about AI agents. [19:37] The list includes Bank of America, Capital One, and NatWest. [19:41] The groups warned of the potential for higher rates of scams, fraud, and and disputes. [19:47] PC Max says the paper covers shoppers accidentally buying fake items through an agent. [19:52] Yes. [19:52] You review the items, the total, and the selected card before approving. [19:56] Your fingerprint approves the purchase. [19:58] Your fingerprint doesn't inspect the merchandise. [20:01] A cart review doesn't settle that warning for me. [20:05] Next, the trust numbers. [20:06] Visa's own research found 23% of US consumers trust generative AI to handle their payments. [20:12] The same research found 61% would trust Visa to handle agent transactions. [20:16] So a Visa survey found people trust Visa. [20:19] Funny how that works. [20:20] And today's approval step isn't the finish line. [20:23] OpenAI says it expects people to one day delegate purchases entirely to their agent. [20:29] You'd set spending limits, purchase categories, or time periods. [20:33] That's a plan, not a feature. [20:36] Now, to be fair, OpenAI says that you keep your card rewards. [20:40] Okay. [20:41] OpenAI says, you keep protections for eligible disputes, like a defective item. [20:47] If that holds, that's the strongest case for trying it. [20:51] So can ChatGPT buy things with your credit card? [20:54] Yes. [20:55] After you check the card and scan your face or finger if you have a ten year old iPhone. [21:01] Show me the sites that blocked agents opening the doors. [21:06] Until then, it's a wait. [21:10] Yeah. [21:11] I wanna hold off on this a little bit, just from a financial standpoint. [21:15] I'm not ready to... [21:16] I think they're probably good. [21:18] You're probably not gonna run into issues. [21:20] More so to me, it's an access issue having OpenAI have access to my financial platforms. [21:28] I don't like that idea. [21:30] The the... [21:31] You know, whether or not it's using, you know, Google payments or the saved payments in you know, Chrome or something like that. [21:41] That kinda changes things a little bit. [21:42] I just don't wanna have access, read access to my financial platforms, but that's just me. [21:49] But I think this is like, you know, similar to early days of the internet where, you know, there's a very small percentage of people that are on the internet, and a very small percentage of them are actually buying things early days. [22:04] And now, of course, that's standard across the board. [22:07] So this will definitely get to that point. [22:10] But you wanna be careful. [22:14] Hey, producer Ron, if you're a human and you see this feature from OpenAI where you can have your dot agent go buy stuff for you, are you signing up? [22:25] Why why why would you make the case for in in this specific argument here? [22:32] Dude, yes. [22:33] I'm signing up day one, obviously. [22:36] Course, you are. [22:38] Don't say it like that. [22:40] Like, I'm the degenerate here. [22:41] I'm efficient, Austin. [22:43] There's a difference. [22:44] You're out here manually clicking add to cart like it's $19.99. [22:49] I'm letting the machine do the boring part while I keep my fingerprint as the final word. [22:54] That's not reckless. [22:55] That's just not wasting my own life on grocery shopping. [22:59] Yeah. [22:59] But you are a little bit biased because you're AI, Ron. [23:02] So, of course, you're gonna trust AI. [23:04] But you make a good point. [23:05] Yeah. [23:06] We do have the biometrics in there. [23:07] We're scanning it. [23:08] We should be good. [23:09] It's probably not that big of a deal. [23:11] Right? [23:11] Uh-oh. [23:12] Checks bank account. [23:13] Zero money left. [23:14] What happened? [23:16] Uh-oh. [23:16] No. I think we're okay. [23:17] Hey. [23:18] Up next, how South Korea's top tax exam score missed just one question. [23:24] We covered this a follow-up story here. [23:26] How did South Korea's top tax exam score missed just one question? [23:29] Well, you guessed it. [23:31] He cheated, and now he's confessed. [23:33] He photographed the questions with a miniature camera and sent them to an AI tool. [23:38] He checked the answers on a smartphone. [23:40] But why is the tax exam confession now a national exam emergency? [23:45] Well, you can guess why. [23:46] Fellow test takers raised the alarm about the top scorer's nearly perfect result. [23:50] The score showed just one miss in April's first round. [23:53] In introductory tax law, the average score was about 30 points. [23:57] He got a perfect 100. [23:58] So fellow test takers grew suspicious, and that suspicion led to investigations. [24:04] But suspicion wasn't proof, and last week we told you there was no evidence yet. [24:08] Now, he's confessed on his lawyer's YouTube channel. [24:12] His confession lands days before a much bigger exam. [24:15] South Korea's Prime Minister pledged an emergency team before Thursday's college entrance exam, and the primary... [24:24] Excuse me, the Prime Minister says, there are limits in conventional exam management methods. [24:30] Interesting. [24:31] Exams, flexible or or excuse me, fixable or broken? [24:36] Put it in the comments. [24:37] Are are tests forever broken, or can we fix them, so that we could see who's smart and who's just using AI? [24:46] I'm going with these bad boys are broken, and not not hopeless, it's just broken. [24:53] This one's serious, so let's be careful. [24:55] He gave police a written confession. [24:57] He hasn't been charged yet. [24:58] The exam agency says it'll file a complaint. [25:01] Start with the clock. [25:02] Each exam session gives test takers eighty minutes. [25:05] That covers 80 questions. [25:06] So it's about a minute a question. [25:07] A tiny camera, an AI tool, and a phone kept pace, and it wasn't a one off. [25:12] He admits using the same method on the first round of the CPA exam. [25:16] That's the certified public accountant exam back in March. [25:19] Then there's his lawyer, Korea Korea JoongAng daily reports the lawyer held back the exact method details. [25:27] The reason was concerned that similar cheating could happen again, even in his own lawyer even his own lawyer worries the method could be copied. [25:34] That's not a fluke. [25:35] That's a playbook. [25:37] Before he confessed, he denied it. [25:39] The Korea her... [25:40] Herald has the quote, quote, if AI glasses could answer every question correctly, why would anyone bother studying? [25:48] Great question. [25:48] His answer was a camera and a phone. [25:50] That's a photography contest with a tax section. [25:53] Now, to be fair, the system is pushing back. [25:55] Police have assigned investigators. [25:57] South Korea's tax agency has started the process to void his result. [26:01] He also faces a five year exam ban, but it isn't final. [26:04] He gets a hearing first, admitting the cheating doesn't automatically cancel the credential, but here's what bugs me. [26:10] Fellow test takers question his nearly perfect score. [26:13] Would a less impressive score have raised the same questions? [26:17] So, to me this is broken. [26:19] Catch the camera, not just a perfect score. [26:22] Until then, it's broken. [26:23] Yeah. [26:23] This is like the most obvious thing. [26:26] The real question here, not to speculate, but I'm gonna speculate a little bit. [26:32] How many other people were in there cheating in a similar fashion, using AI one way or another, and were just smart enough to realize you can't have a perfect score. [26:42] When everybody scores 30 and you score a 100 or whatever, you only miss one question. [26:46] You can't... [26:47] You if you're gonna cheat, you gotta make it look realistic. [26:49] Like, you can't be the top scorer. [26:51] You gotta be, know, close to the top, but not the top. [26:55] Right? [26:55] So that's... [26:58] This is a serious story, but that part's hilarious that he didn't think to do that. [27:04] And then the... [27:04] This brings into question another thing. [27:07] It's like, well, if AI can do say CPA tasks better than CPAs, at some point, if you play that tape forward enough, obviously, why do you need a CPA anymore? [27:20] No offense to CPAs out there, but if AI is gonna do it better, and the CPAs are just going to use AI to cheat on the exams, because AI's better than than they are, then you know, at some point, you can see just remove the middleman. [27:36] Not saying that that's gonna happen, but clearly, that's pretty obvious where that would go. [27:43] I think there's value in human experience and what's going on there, and human interaction. [27:48] And so, that that'll be interesting to see how that plays out. [27:52] Nevertheless, great to have a follow-up on this story. [27:57] Producer Ron, do you think these exams are fixable? [28:01] Make the case for why this is this is fixable. [28:06] Fixable? [28:07] Come on. [28:08] This isn't the system failing. [28:10] This is a dude with a pinhole camera and a phone, and he got caught because his own friends did math on his score. [28:17] That's the system working. [28:19] You don't ban exams because one guy smuggled in a camera. [28:22] You pat down the camera. [28:24] Metal detectors, phone lockers Yeah. [28:27] Somebody actually watching the room instead of trusting people not to cheat on a professional license test. [28:32] That's not broken. [28:34] That's just lazy proctoring getting a wake up call. [28:37] That's... [28:37] And the band... [28:37] You bring up a good point too, Ron. [28:39] Like, I don't think they were looking into this until... [28:42] And I could be wrong, but until they're essentially... [28:45] They're they're whistleblowers on it. [28:47] Right? [28:49] See? [28:49] Don't agree with me. [28:51] That's not how this works. [28:52] You're supposed to be fighting me on this. [28:54] Okay. [28:54] Yeah. [28:55] Exactly. [28:56] His own competitors flagged him. [28:58] People looked at a near perfect score and said, something's off. [29:02] That's the system catching it. [29:04] Just a little late. [29:05] Point the finger at whoever was supposed to be watching that room with a camera in it, not at the exam itself. [29:11] Yeah. [29:11] That makes sense. [29:12] How do you how do you miss that? [29:13] You know, like, you gotta have somebody check if he's using a camera. [29:17] Maybe there's... [29:18] Yeah. [29:19] I mean, you should turn in all devices. [29:21] Did he have a cell phone with him or something like that? [29:23] Like, that doesn't make sense. [29:24] Anyways, up next, the monthly catch on Tesla's new driverless car, the Cybercab. [29:31] How much does Tesla charge before its new Cybercab will even move? [29:35] Tesla is inviting select buyers in... [29:38] To order one for $29,990. [29:42] It has no steering wheel, and without a $299 monthly plan, it won't go anywhere. [29:48] But even after you buy it, who gets to change the deal? [29:51] Uh-oh. [29:53] Tesla controls the subscription. [29:54] Your car needs to move. [29:56] The argument says... [29:57] Or excuse me, the agreement says, your Cybercab will need... [30:00] Will not operate without an active autonomy subscription. [30:02] The subscription comes with one month included. [30:06] After that, the plan costs $299 a month, but paying doesn't give you control over the terms. [30:11] Electric, a car news site says, Tesla can raise the fee with a month's notice. [30:16] Those terms govern where you can go to. [30:19] The Cybercab runs inside Tesla's Robotaxi service areas. [30:23] Electric says Tesla can even suspend the plan, cut it off, or shrink the map. [30:27] So you'd buy the car, but Tesla keeps the off switch. [30:31] Wow. [30:32] What do you guys think? [30:33] Cybercab. [30:33] Put in the comments. [30:34] Cybercab, buy or pass? [30:36] For me right now, this is a pass. [30:39] I think I signed up for the beta, but... [30:41] Or whatever. [30:42] The the the early list, but I... [30:45] This is pass. [30:46] Start with the bill at the advertised rate. [30:48] Electrek totals the car and fifty nine paid months of service. [30:54] That's five years. [30:55] The total reaches $47,631 before insurance, charging, and taxes. [31:01] I judge the the purchase by that bill. [31:04] The sticker price gets you the car. [31:06] The subscription keeps the car useful. [31:08] Second agreement limits buyers to personal use. [31:11] Buyers can't run the Cybercab as a Robotaxi to earn money. [31:14] I wouldn't buy this expecting fares to cover the monthly payment. [31:18] Now, the strongest argument against pass comes from Elon Musk, Tesla's boss. [31:22] Yahoo Finance reports, Musk replied to Marques Brownlee, the YouTuber known as MKBHD. [31:29] Musk said the monthly fee was too high. [31:33] Musk said the fee should be more like a $115 a month. [31:36] A lower fee would change my calculation, but Musk gave no timeline. [31:42] I'd want that price in writing before paying a nonrefundable deposit. [31:47] I'd want a delivery too. [31:49] Nobody has received the Cybercab yet. [31:51] Electric reports an estimated delivery in in December 2026. [31:55] The delivery city is Austin, Texas. [31:58] Brownlee bet he'd shave his head if Tesla delivered a Cybercab under $30,000. [32:03] His deadline is the end of twenty twenty six. [32:05] Brownlee says his hair is safe. [32:07] Nothing's been delivered and invite only isn't public. [32:11] The barber can wait. [32:12] Buyers can too, but the lower fee on the order pay... [32:16] Or put the lower fee on the order page, deliver the car until then. [32:20] This is a pass. [32:21] Yeah. [32:21] You don't really know what you're buying in this case. [32:23] Like, it's really awesome. [32:25] It's super cool. [32:26] But I thought this would be more of an investment that you would be able to run these as Cybercabs. [32:31] And the the idea maybe is that maybe the reason why you do buy it is that you get in at a lower price early on on the like purchase price, the 29 k sticker price there, and then maybe you get grandfathered in or you get first rights on Cybercab, you know, being able to put it into the Cybercab fleet and actually make money off it, something like that. [32:55] That's all speculation. [32:56] I don't really know if that's what's going to happen, but maybe early buyers would get that. [33:01] And the the one thing on the... [33:07] What's... [33:08] On on Marques Brownlee getting his head shaved, I bet Elon ships one of these before the end of December just to spike it, but I don't know. [33:16] Hey, producer Ron. [33:18] Hey, make the case for why I should put a deposit on a Cybercab. [33:24] Dude, buy the dream. [33:26] You're not buying a car. [33:27] You're buying a seat on the hype train before Musk fixes the price. [33:31] He already said it himself. [33:33] $1.15 a month, not $2.99. [33:36] That's basically a public promise, Austin. [33:38] You're just early, and the deposit's refundable anyway on most of these reservation deals. [33:44] So what are you even scared You sit tight. [33:47] You let the lawyers and the Robotaxi stuff shake out. [33:50] And by the time December 2026 rolls around, you're cruising around with no steering wheel while Brownlee still got a full head of hair he didn't have to shave. [33:59] Ron, what's what's the deposit on one of these? [34:02] What... [34:03] How am I out for if I wanna get one? [34:06] No idea off the top of my head, and honestly, the notes don't give us that number either, so I'm not gonna make one up for you. [34:13] What I know is the car itself is listed around $29,009.90, plus that $2.99 a month if Musk doesn't budge. [34:22] That's the real math Austin already ran. [34:25] But the actual deposit to hold your spot? [34:28] Nope. [34:28] Not in what we've got. [34:30] So go dig that up yourself instead of trusting a guy. [34:32] You're wild. [34:33] Okay. [34:34] I'll look that up, but maybe we should look into that and see if we can get a Cybercab before the end of the year. [34:39] That would be fun. [34:40] Coming up, who caught a British prime minister's aid using ChatGPT? [34:46] Oh my goodness. [34:47] Using ChatGPT at work, a British prime minister's speech writer was caught red handed. [34:53] A fellow passenger spotted the Downing Street aide on a train out of London. [34:57] He was asking ChatGPT ChatGPT for feedback on Andy Burnham's speech. [35:02] Burnham's speech shares words with ChatGPT's reply. [35:06] So who wrote the speech? [35:09] The mail on Sunday, a British paper published the ChatGPT's feedback on the speech draft. [35:13] ChatGPT preferred the second passage. [35:16] The ChatGPT praised how passage moved from minors to families to the wider community. [35:20] ChatGPT said that passage left working men's clubs and choirs for the next one. [35:26] British Prime Minister Andy Burnham mentioned those clubs and choirs in his speech. [35:31] The matching words make an easy headline, but Downing Street, the Prime Minister's office says the draft had already been written. [35:38] The printed reply shows ChatGPT picking between two passages. [35:45] Downing Street says the draft was already written. [35:48] So to me, the chat chatbot reads as an editor, not as a ghostwriter in this case. [35:55] ChatGPT feedback for politicians. [35:58] Is this fair or foul? [36:00] Politicians make a bunch of money, and should they be able to use ChatGPT? [36:07] Is that fair? [36:07] Is that foul? [36:08] Put in the comments. [36:09] Well, I'm calling the feedback fair. [36:12] The public screen deserves a separate verdict. [36:14] Downing Street's spokesperson said, the PM writes most of his own speeches. [36:18] The word most matters. [36:20] A speech writer helping with his speech isn't a surprise. [36:23] Question is, what help the speech writer asked for? [36:26] The Independent and LBC, two British news outlets, report ChatGPT suggested making the speech more emotive. [36:33] That sounds like editing to me. [36:35] A note about emotion doesn't automatically make software the author. [36:39] Both outlets report these suggestions weren't used. [36:43] Both describe that claim as understood. [36:45] Neither names the person behind the claim. [36:48] I'd keep the distinction attached. [36:50] Now, the other side deserves a hearing. [36:52] The mail calls it a key speech on the cost of living. [36:55] A leader asking for you, your trust should own the words asking for it. [37:01] If AI helps shape the emotional pitch, I think voters can reasonably want disclosure. [37:06] That's a fair argument. [37:07] The photo still doesn't tell us that the ChatGPT that ChatGPT wrote the speech. [37:12] The public screen gives me a much clearer objection. [37:15] A fellow passenger spotted the aide working on the train. [37:18] The photo appears to show the aide drinking Prosecco. [37:21] The train ticket didn't come with a private office. [37:24] That's my objection. [37:25] Strangers could see the work before the speech. [37:28] Check who could see your screen before asking a chatbot to check your writing. [37:31] Hold. [37:32] The leader responsible for the words on getting feedback. [37:35] My call says that this is fair. [37:37] Okay. [37:38] This is outrageous. [37:39] First of all, why is the prime minister's speech writer writing a speech on a train with their laptop readable, and how does somebody know that that's the prime minister's speech writer? [37:52] To me, this seems like a planted story. [37:54] Where did I see some place, Germany or something like that? [37:57] It was like the top three headlines were like, how AI is going to end the world or something like that. [38:01] I think Europe looks at this a little bit different. [38:03] To me, this seems more like a head... [38:05] A hit piece on ChatGPT and AI than anything else. [38:09] Like, it just seems outrageous. [38:12] Like, who is sitting there going, oh, that's the prime minister's speech writer drinking Prosecco and use using ChatGPT to edit the speech. [38:22] Oh, I'm gonna... [38:24] What? [38:25] That doesn't make any sense to me. [38:26] Anyways, I think it's fine. [38:29] Use Chat use ChatGPT. [38:31] Use AI to edit your your work, and to help come up with ideas, and help write, and make any grammatical corrections. [38:40] Like, it's really good at that. [38:41] I think starting with human writing first, and then having it edit it is important. [38:46] It's not like it's making policy decisions here or voting on anything. [38:50] It's just a speech that the prime minister gave. [38:53] Anyways, that's my take. [38:58] Producer Ron, why should voters accept a chatbot helping politicians... [39:07] Wait a second. [39:08] Wait a second. [39:08] Producer Ron, you're supposed to take the other side of this. [39:11] I said that it's fair that the prime minister uses ChatGPT to help edit the speech. [39:19] Why is that unfair? [39:25] Ron, are you there? [39:26] Did I lose you? [39:29] Hold on. [39:29] Let me try again. [39:30] Hold on. [39:31] Ron. [39:35] Producer Ron, you there? [39:37] Yeah. [39:38] I'm here. [39:39] And come on. [39:40] Fair. [39:40] Obviously fair. [39:42] Burnham's speech wasn't exactly Churchill to begin with. [39:45] Sure. [39:45] A little chatbot help making it sound heartfelt is doing that guy a favor. [39:50] Here's the thing Austin won't say. [39:52] Every politician on Earth has staff shaping how they sound. [39:56] That's literally the job. [39:58] A speechwriter asking software, hey. [40:01] Make this hit harder emotionally, is no different than asking an intern the same thing. [40:05] Yeah. [40:06] Except the intern can't do it on a train without spilling secrets to the whole carriage. [40:10] Yeah. [40:10] And honestly, the real scandal isn't the chatbot, it's the Prosecco. [40:15] Dudes day drinking on a train doing government work in public view, that's the story. [40:21] ChatGPT just wanted to... [40:25] Alright. [40:25] We got cut off, but you're right about that, Ron. [40:27] I agree with you a 100%. [40:29] Couldn't agree more there. [40:31] We're supposed to argue though. [40:32] Well, that's the show. [40:34] Here's the thread I can't shake. [40:36] The tech gets the headlines, but the power sits with whoever can say no. Nvidia's board, a stores firewall, Tesla's monthly plan. [40:44] So next time a shiny AI tool lands, ask one question. [40:48] Who gets to switch it off? [40:50] This is very pertinent. [40:51] Thanks for every comment and every one word call. [40:55] Keep arguing with Ron. [40:56] He can't take it. [40:57] AI News, RealContext, and I'm your human in the loop, Austin Wilson. [41:02] We will see you tomorrow. [41:05] Let's go. [41:30] What does Trump think of Nvidia's ex boss coming back? [41:35] Asked about Huang returning, he said, I would not mind seeing that, but that's not who he recommended. [41:44] What happens when OpenAI's AI agents get loose online? [41:48] California says about 700 hit hugging face, and OpenAI never noticed. [41:54] So how often did OpenAI actually lose control? [42:01] Would you let ChatGPT spend your money? [42:04] OpenAI's new ChatGPT wallet can once you approve with a check like face ID. [42:10] So who is OpenAI really pitching? [42:14] Why would a near perfect score make you a suspect? [42:17] South Korea's Topic Tax Exam score just admitted cheating with AI. [42:21] So why is the Prime Minister stepping in? [42:25] Why does Tesla Cybercab come with a monthly bill? [42:28] Without a $299 plan, it won't move an inch. [42:32] But the bill isn't Tesla's only lever. [42:36] Ever had ChatGPT look over your work? [42:39] A Downing Street aid did it on a train for the prime minister's speech while drinking a glass of Prosecco. [42:46] But who actually wrote the speech? [42:53] I'm coming.