The Most Powerful AI Ever Built, Cybersecurity Risks, and What It Means for You

On April 7th, Anthropic announced Claude Mythos — described as “by far the most powerful AI model we’ve ever developed.” But instead of releasing it to the public, they locked it away. In this episode, Ginny explains what Mythos can do, why it matters to everyday people, and what the emergency meeting between the U.S. Treasury, the Federal Reserve, and America’s biggest banks was all about.

Mythos identified thousands of zero-day vulnerabilities — hidden security flaws — across every major operating system and browser, including a 27-year-old bug in software used in internet routers. Anthropic responded by forming Project Glasswing, a coalition of 40 organizations working to patch those vulnerabilities before bad actors can exploit them. Ginny breaks it all down in plain English, including what you can do right now to protect yourself.

Also in this episode — ten hands-free Siri commands for safer driving, Palantir CEO Alex Karp’s provocative take on which workers will thrive in an AI economy (hint: it’s not the elite degree holders), plus recommendations for the NYT Cooking app and a five-ingredient lemon-lime cream dessert that looks far more impressive than it is.

Links to all articles and the recipe are in the show notes. And don’t skip this week’s homework: check for software updates and install them. After this episode, you’ll know exactly why.

SHOW LINKS:

10 Things You Can Ask Siri While Driving — https://medium.com/@justinpineda/10-things-you-can-ask-siri-while-driving-250595c8f2e7

Palantir CEO Says Only Two Types Will Survive AI — https://medium.com/predict/palantir-ceo-says-only-two-types-will-survive-ai-and-elite-degrees-arent-one-of-them-341c222044e0

Lemon-Lime Cream Recipe — NYT Cooking - https://www.nytimes.com/subscription/cooking.html

CHAPTERS:

00:00 Welcome to aiGED

01:09 Where the Stories Come From

02:35 Siri Tips for Driving

04:10 Palantir CEO on Jobs

07:02 Introducing Claude Mythos

09:02 Cybersecurity Basics

11:18 Mythos Finds Zero Days

13:01 Project Glass Wing Access

14:18 Government and Bank Alarm

16:44 Why Anthropic Locked It

17:18 What It Means for You

20:11 Apps and Dessert Picks

22:22 Homework and Wrap Up

aiGED: AI for the 65+ crowd

Transcript

SPEAKER_00 0:00

Well, hello everybody, and welcome to Aged, the one and the only podcast that is devoted to AI and the 65 Plus crowd. I'm your host, Jenny Deeran, and today I want to talk about what I think is the most important AI story of the year, maybe the decade. It's about the most powerful AI model ever built and why the company that made it decided to lock it away instead of releasing it to the world. I'll tell you what it can do and why it matters to every single one of us. It's a big story. We've also got 10 clever things you can ask Siri while you're driving. A CEO who says elite college degrees are basically worthless in an AI economy, which is either alarming or liberating, depending on how you look at it. And a five-ingredient dessert recipe that will make you look like a genius. So let's get into it. Starting with AI in the news, and before I get into the stories, I want to tell you about where I found them. A publication called Medium. Medium is an online platform where writers, journalists, researchers, and everyday experts publish articles on just about every topic you can imagine. It's not a newspaper, it's not a blog, it's kind of in between. Think of it as a really smart magazine that lives entirely on the internet. Barack Obama has published there, Ariana Huffington writes there. So does Evie Williams, who actually co-founded Medium, and he also co-founded Twitter, which obviously has gone the ways of Elon Musk. The articles that rise to the top tend to be thoughtful, well researched, and written by people who really know their subject. You can read a handful of articles for free each month or go unlimited with a membership. It's about $5 a month or $50 a year, and I found it to be a great place to discover AI stories and other stories that go a little deeper than the typical headlines. You can find it at medium.com and I highly recommend you check it out. So, story number one that came from the Medium: 10 Things You Can Ask Siri While Driving. My first story was written by Justin Paneda and it's called 10 Things You Can Ask Siri While Driving. It was published on Medium. Now, I know a lot of us already use Siri, but I'll bet most of us are only scratching the surface. Justin lays out 10 genuinely useful things you can do completely hands-free while you're behind the wheel, and some of them I had never even thought of. The obvious ones are there. Ask Siri to make a call, send a text, or get directions to the nearest gas station. Those are all pretty common. But here are a couple that surprise me. You can ask Siri to set a location-based reminder, like remind me to pick up bread when I go to the grocery store. It will actually ping you when you arrive. And you can ask Siri to turn on Do Not Disturb while you're driving, so you're not tempted to glance at your phone every time something buzzes. My personal favorite from the list, you can ask Siri to play a podcast at one and a half times speed, which means you could be listening to Aged in the car at regular speed, or you could be a speed listener. Your call. The point is, Siri is more capable than most of us realize, and using it while driving is genuinely safer than reaching for your phone. Worth a read. The writer is Tasmia Charman, published on the Medium, as I said, on March 27th of this year. Now, full disclosure, I find Palantir to be a pretty unsettling company, to say the least. They work deep inside U.S. intelligence and military operations, and their leadership doesn't exactly scream careful and responsible. But their CEO, Alex Carp, makes some points that I think are genuinely worth thinking about. Palantir is a massive data analytics company. Think a $433 billion company. And CARP has a blunt message for anyone who went to fancy schools and is counting on that degree to protect them from AI. He says, and I'm quoting, there are basically two ways to know you have a future. One, you have some vocational training. Or two, you're neurodivergent. So electricians, plumbers, technicians, people building the physical infrastructure that AI runs on, those jobs are hard to automate and in sky-high demand right now. And neurodivergent thinkers, people with ADHD, dyslexia, autism, who've spent their whole lives having to solve problems sideways because the playbook never worked for them. Carp says that kind of creative, unconventional thinking is exactly what an AI economy rewards. To put his money where his mouth is, Palantir launched a neurodivergent fellowship, paying up to $200,000 a year. It drew 2,000 applications in days. Now, not everyone agrees with CARP. Dario Amade, co-founder and CEO of Anthropic, the company that makes Claude, my AI, says the opposite. He believes qualities like emotional intelligence, communication, kindness, and curiosity will become more essential, not less, in an AI world. My take, both of them are probably a little bit right, but it's a fascinating conversation and one I think is really worth paying attention to. So I'll put links to both of these articles in the show notes for this episode. Now let's talk about the main topic for this episode, and that is Claude Mythos, the most powerful AI ever built, and why you need to understand it. Okay, I meant what I said in the intro. This is a big deal, and I promise I'm going to explain it in plain English, because this is exactly the kind of story that gets buried in tech jargon and never makes it to the people who actually need to understand it. That's you, that's us. So let's start at the beginning. What just happened? Well, a few days ago, on April 7th, Anthropic, the company that makes Claude, the AI that I use on this show, announced a new AI model. It's called Claude Mythos. Claude Mythos. And here's the thing: they announced it, but they are not releasing it to the public. They are deliberately keeping it locked up. And the reason why is what makes this story so important. Anthropic says Claude Mythos is, in their own words, quote, by far the most powerful AI model we've ever developed, close quote. It is so capable and potentially so dangerous that they decided not to let anyone outside a very small, very carefully chosen group of companies get their hands on it. So just a very small group of people it's being released to. Now, when a company builds something and then says, we're not going to sell it, you've got to ask why. What is it capable that scared them enough to do that, to keep it under lock and key, except for this small group? I'm going to tell you, but first, let me give you a little background. Because to really understand why mythos matters, we need to understand more about cybersecurity. Here's a quick plain English primer on cybersecurity. Think of your house. Your house has doors and windows. Most of them are locked. But maybe there's a window in the back that doesn't latch properly. You don't know about it, you've lived there for years, and nothing has happened. But if a burglar went around testing every window in every house on your street, they might find it. Computer systems work the same way. Every piece of software, your phone's operating system, your web browser, the programs that run your bank, your hospital, your electric grid, has what are called vulnerabilities, hidden weaknesses. Flaws in the code that were never caught. Broken window latches never noticed. Some of these flaws have been sitting there for decades. Nobody found them, not because people weren't looking, but because there are billions of lines of code in the world, and it takes an enormous amount of human time and expertise to go through it all. Here's a term worth knowing: zero day vulnerability. It sounds technical, but the concept is simple. It's a hidden flaw that nobody has found yet. Zero days means you've had zero days to fix it, because as far as you know, there's nothing to fix until there is. Finding zero day vulnerabilities, the good guys use them to patch software before it gets attacked. The bad guys use them to break in. Until now, finding these hidden flaws required expert human researchers, specialized tools, and a lot, a lot, a lot of time. That's changing very fast. So what Mythos can do. In just the past few weeks, Anthropic used Claude Mythos in their lab environments to identify thousands and thousands of zero-day vulnerabilities across every major operating system and every major web browser. We're talking about Windows, Mac, Linux, Chrome, Safari, the software that runs your phone, your tablet, your laptop. One of those identified vulnerabilities was a twenty-seven-year-old bug in OpenBSD, an operating system used in internet routers and secure firewalls that would allow a hacker to remotely crash any machine running it. Twenty-seven years it was sitting there hidden, and Mythos found it. Here's the metaphor I find helpful. Imagine a burglar who can check every window and door lock in every home in America simultaneously in a matter of minutes. Not just your neighborhood, every home, every building, every system. That's what we're talking about. Here's the hard truth. If this technology can be used by the good guys to find and fix those weaknesses, it can also be used by the bad guys to find and exploit them. That's what makes this a two-sided story. So who has access to mythos right now? Anthropic formed something called Project Glasswing. Glasswing. A coalition of about 40 organizations, including Amazon, Apple, Microsoft, Google, JP Morgan Chase, CrowdStrike, and the Linux Foundation, allowing them to use Mythos Preview specifically for defensive security work. The name Project Glasswing comes from the Glasswing Butterfly, which I'm not familiar with. But it's a creature with transparent wings that hides in plain sight. The idea is that the world's most dangerous software vulnerabilities have been hiding in plain sight for years, buried in complex code that no human ever fully searched. Mythos can finally see them. The goal is to find the weaknesses and fix them before the bad guys find them and use them. It's basically a race. And Anthropic is trying to make sure the good guys have a head start. And then the government got involved. Now, if the story ended there, a powerful AI model, a responsible company, and a defensive coalition, that would already be significant. But this week it got even bigger. On April 10th, the New York Times reported that Treasury Secretary Scott Bessent called a hastily arranged emergency meeting in Washington. In the room, the chief executives of some of America's largest banks, Bank of America, City, Wells Fargo. Also in the room, Federal Reserve Chair Jerome Powell. Think about that for a second. The head of the U.S. Treasury, the head of the Federal Reserve, the leaders of the biggest banks in America all gathered on short notice to talk about one thing, Claude Mythos. The message delivered at that meeting was blunt. The new model might be so effective at finding security weaknesses inside banks that if hackers or any bad actor got their hands on it, they could use the same capability to exploit those weaknesses rather than fix them. Your bank account, your savings, the financial system that underpins the entire economy, all potentially at risk. Now, here's something I find genuinely remarkable about this. The Trump administration and Anthropic are currently in a legal battle. You may remember from episode 31, the Pentagon designated Anthropic as a supply chain risk because Anthropic refused to remove safety limits on how its AI could be used in war. They are fighting in court right now. And yet, even with that fight going on, Kevin Hassett, the director of the National Economic Council, went on Fox News and said this quote, we're taking every step we can to make sure that everybody is safe from these potential risks, including Anthropic agreeing to hold back the public release of the model until our officials have figured everything out. There's definitely a sense of urgency. So why did Anthropic make this choice? I want to pause here for a moment because I think the decision Anthropic made is as important as the technology itself. They could have released mythos to the public. It would have made the money, probably a lot of money. Instead, they announced it, explained why it's dangerous, kept it locked down, and launched a defensive coalition. That sounds beyond responsible to me. So what does this mean for you and me? Now, I know what some of you might be thinking, Ginny, I'm not running a power grid, I'm not operating a bank. Why does this affect me? Well, here's why. The software that runs your phone, that's the same software this affects. The systems that protect your bank account, the medical records your doctor's offices keep, the infrastructure that powers your neighborhood. I've noticed I'm getting much more regular software upgrade alerts for Google, Microsoft Outlook, and of course Claude. This is not hypothetical. The attacks are already happening. AI is already being used against us. Mythos is Anthropic's attempt to get ahead of that, to arm the defenders before the attackers figure out how to use this kind of power on their own. As one cybersecurity analyst put it, there will be more attacks, faster attacks, and more sophisticated attacks. Now is the time to modernize. The honest bottom line? I'm not going to tell you this isn't frightening because I think it is, but I also want to give you the full picture. The fact that Anthropic found thousands of vulnerabilities in the world's most important software, that's actually good news. If those vulnerabilities get patched before anyone else finds them, and that's exactly what Project Glasswing is designed to do, Anthropic said that the same capabilities that make AI dangerous in the wrong hands make it invaluable for finding and fixing flaws and for producing new software with far fewer security bugs in the first place. Think about it this way: you'd rather have a very thorough home inspector find every crack and weakness in your house before you move in than discover them after the storm hits. That's what Mythos is doing for the world's digital infrastructure. The arms race between hackers and defenders has been going on as long as computers have existed. What's changed is the speed. Of course, it's the speed. And now, for the first time, the defenders have a tool that might actually help them keep up. I don't think any of us should panic, but I do think all of us should pay attention because AI is no longer just about chatbots and homework helpers. It is now operating at a level that touches the systems all of us depend on every day. And that, my friends, is a very big deal. So let's get a little lighter and get into my recommendations. My first recommendation is the New York Times Cooking App. If you love to cook, or even if you just love to eat, this one is worth it. It's the New York Times dedicated recipe platform, and it is beautifully done. We're talking tens of thousands of tested recipes, everything from quick weeknight dinners to showstopper holiday dishes. You can save recipes, you can make grocery lists right inside the app. You can also watch how-to videos and get personalized suggestions based on what we like. So you can type by the ingredients you already have. It's really a handy app. It's about $5 a month or $40 a year. Check the prices on the New York Times site. But I highly recommend this. And my second recommendation is actually linked to that New York Times cooking app. I'm sharing it because it is almost embarrassingly simple for how impressive the dessert looks and tastes. It's a lemon lime cream, a silky custardy dessert you bake in little ramekins. Five ingredients: lemon, lime, sugar, eggs, and heavy cream. That's it. I added a little ginger and I put some lavender olive oil, little drops on the top of it. Use your imagination. Put those cups in the little baking pan. Pan and pour some water in it about halfway up the cups and bake it slowly for about 30 minutes and that's it. It is delicious. You can serve it chilled or at room temp. And people are going to think you spent a lot of time cooking it, and it's super great. So I highly recommend that. Now for this week's homework, check your software updates on your phone, on your tablet, on your computer, and install them. I know we all hit remind me later, but after everything we've talked about, after all the information about mythos and finding all these vulnerabilities and how all these companies are going to be updating their software as they fix these vulnerabilities. So please pay attention to them. They're often exactly the kind of security patches that we need. So it takes about 10 minutes to update your software, and it genuinely matters. So that is your homework assignment. So let's see, before we wrap, I want to remind all of us that I mean this episode has been a lot about the hazards of AI. But we all try to focus a lot on the helpful side. That's the side that we live on at Aged, but I don't want to ignore the hazardous side. So that's why I've taken this whole episode to really talk about some of the current challenges. But on your end, be sure that you protect your information, double check advice, and trust your own judgment. And if you love listening to this podcast, please pass it on to your pals, especially in the 65 plus crowd. And thank you for listening. And remember, it is never too late to learn something new, especially something that might make life easier and especially more fun. Cheers.