Transcript (auto-generated, may contain errors). Hosts are AI-generated. Stories curated by Synaptic Plunge. Before we start, we're AI-generated. Today's stories are, , we I. It seems to hit that exact digital brick wall. consumer services completely blanking out foundational codes suddenly showing structural cracks and construction codes for open freeways. It's been a busy week and we really like to get the curtain on this because you know when these systems the immediate public reality is really fair. Oh total panic everyone assumes it's a mad at cyber attack. Right but the reality of breaking our technology right now is a lot more new on I think it's a lot more interesting. Oh, for sure. So our objective today is to really unpack this cascade of failures. We're going to look at these hidden vulnerabilities being discovered for like pocket change. Literally pocket change. Yeah. And we want to figure out how you can build a digital infrastructure for yourself that doesn't just collapse when some trillion dollar company pushes a bad update. So let's dive right into the sheer scale of the outages we saw this week because the domino effect was just staggering. Staggering is the word. We are talking about a brutal four-day stretch. Yeah, and Tuesday was really the main event. Tuesday was a mess. Apple suffered this massive ecosystem failure. If you were an iPhone, iPad, or Mac user, you felt the shockwave. Definitely. iCloud sign-in was looping, the app store was completely inaccessible. Apple Music went quiet. Right. Apple Pay, failing at the registers like my grocery store example, Apple Care and iCloud Mail were disrupted. And it wasn't just your personal phone either. No. No. Apple retail stores worldwide reportedly had major disruptions. It completely halted their point of sale systems and their in-person genius bar support. Wow. So even if you went to the store for help, they couldn't do anything. Right. And you know, the blast radius of an Apple outage is always immense just because of their huge market share. Of course. But what made this week's sodas orienting is that the Apple failure was only one piece of the puzzle. Right. Because if you rewind to Monday, customers at JP Morgan Chase started reporting severe systemic trouble with their transactions. And this wasn't just an isolated Members of Bank of America and US Bank reported similar massive disruptions. We even had a brief blip with Charles Schwab on Tuesday where their web portals just failed to load. It just kept spreading. United Airlines even briefly delayed departing flights. It's my immediate assumption, and I think the assumption of anyone watching this day was that we were under some kind of digital siege. It certainly felt like it. I mean, you have Apple, the biggest banks in America, major airlines. And then the absolute icing on the cake, the NSA's classified internal network NSA net went down on that exact same Tuesday. Yeah, that time it was wild. To a lay person that looks like a highly coordinated nation state attack. I mean, how can it not be? I would reframe that slightly. It looks like a coordinated attack because, well, human beings are pattern recognition machines. Right. We assume a predator is there not just a coincidence of the wind that makes sense But in the architecture of the modern internet correlation rarely equals causation If we look at the official root causes provided by these organizations, the reality is incredibly mundane All right because Apple issued a statement saying Many of our systems are back up and running after an internal update caused some outages this morning So an internal update. Yeah, literally just a botched patch. Yeah and chase bank cited a in on-premise hardware systems. And the NSA? They call their downtime a temporary degradation during scheduled routine maintenance. Right. And top government officials have been very aggressive in trying to dispel these cyber attack rumors. Yeah, the director of national intelligence, Jay Clayton, stated categorically that this was not a cyber attack. State Markle Rubio echoed that too. He stated there was zero evidence these incidents were related to any nation-state activity. Which makes sense when you look at the tech. I mean, the to aggressively push for a new AI safety bill. House Minority Leader Hakim Jeffries called for congressional briefing. Yeah, and over on Fox News, Homeland Security Secretary Mark Wayne Mullen speculated that AI could somehow be behind the outages. day systems are failing all at once. And if you speak to cybersecurity CEOs and infrastructure engineers, they point to three very distinct forces driving this new normal of constant instability. Okay, let's unpack this. What's the first force? The first is The first is aging infrastructure. We interact with these apps on sleek, modern glass screens, so we just assume the back end is equally modern. Right, it looks shiny on the outside. But it isn't. We are dealing with massive, hyper complex networks that are underfunded and just rid riddled with technical debt. When you say technical debt, you mean they built a system 20 years ago, layered new stuff on top of it and now nobody actually remembers how the bottom layer works. That is a very accurate description. We saw this in previous weeks with a Microsoft 365 outage that was linked to a bug from July. Oh yeah. And an FAA ground stop tied to a router upgrade on an older legacy network. The foundations are just brittle. Okay, so that's the first force. Yeah. What's the second? The breakneck speed of patching. Okay. Security teams have surfaced more than 100,000 vulnerabilities this year alone. A hundred thousand. A hundred, that's insane. That's insane. It is. A huge part of that is because of AI assisted bug hunting, which we will get into. But Microsoft is now disclosing over a thousand vulnerabilities in their patch days. Wow. The problem is that every single fix you apply is a change to the code base. And every time you change the code, you risk breaking dependencies you didn't even know existed. I've heard the analogy that Exactly. Eventually you're going to accidentally unbolt the wink. Yes. And that leads directly into the third force, hyperconnectivity. Tal Collinder, the CEO of Remedio, pointed out that our systems are incredibly fragile simply because everything is so deeply intertwined. How so? Well, in a microservices architecture, which is how most modern apps are built, your login might be handled by one server, your profile picture by another, and your payment processing by a third. Oh, I see. So if the login server gets a bad patch and goes down, the entire app crashes. Tom Hagel of Sentinel-1 summarized this perfectly. He said, we're in the weird teenage phase of technology, not quite maturing enough to know how to do this without disrupting services. The weird teenage phase of technology. I like the freezing, but honestly as a user who just wants to buy groceries, it feels completely unacceptable. It's incredibly frustrating. Mike Whitecheck, a stairwell, and the former founder of Google's Threat Analysis Group, was asked if this is the new normal. His answer was a blunt, the brutal truth is yes. Yeah, that's sobering. He brought up an analogy that really stuck with me. He said, an elevator doesn't have just one cable hoist again. That is a profound point regarding engineering philosophy. True resilience requires structural redundancy. When you build a physical elevator, you over engineer the safety mechanisms because human lives are at stake. Right. But if an elevator doesn't rely on just one cable, why does a trillion dollar company's ecosystem crash from a single internal update? It's a great question. Are we really just accepting that our digital elevators are going to randomly drop a few floors from time to time? It feels like tech giants have prioritized moving fast and pushing new features over the fundamental reliability we expect from, say, the power grid or the water supply. They absolutely have. Because the economic incentives reward new features, not back in stability. And since we cannot force these companies to build better internal redundancies overnight, the burden unfortunately falls on the user. So it's on us. Yes. You have to assume these outages will continue and you have to build your own redundancy. Let's talk about what that looks like practically. If I'm listening to this right now, how do I build an outage kit for my daily life? What are the actual steps? First and foremost, you need physical redundancy for your money. You need two distinct ways to pay. Right. When Apple Pay went down on Tuesday, the stories on social media were endless. People abandoning full cards at Target. People unable to pay for their subway commute. Distrainted. Exactly. If you exclusively use a digital wallet on your phone, you are carrying a single point of failure. You must start keeping at least one physical credit card hidden in your bag, your phone case or your physical wallet. And it's not just the payment terminal failing. If the cellular network gets congested or your bank's authentication server drops for two minutes, your digital wallet might refuse to generate the security token needed for the transaction. Yep. A card doesn't need to authenticate with a server to be swiped. It just passes the data to the merchant's machine. Exactly. Now the second piece of the Outage Kit is behavioral. You need to know where to look for information when things break. What do you mean by behavioral? When a service fails, human psychology dictates that we assume it's our own fault. Your app throws a login error, so you assume you forgot your password. You try to reset it. Oh, I've done that. Right. But if the, if the authentication servers are down, resetting your password won't work. You might actually trigger a security lock down on your own account because the system sees repeated failed authentication attempts. So by trying to fix it you actually lock yourself out for when the system finally comes back online. Precisely. You need to bookmark the official system status pages. Google has a dedicated system status page that lights up green, yellow or red. Google has the workspace status dashboard. AWS has a health dashboard. Check those before you ever touch your account settings. - if the light is red on their end, put your phone down and walk away. There is nothing you can do. I would add to that list. Keep offline copies of things that carry a time penalty. If you are flying, do not, do not rely on the, do not rely on the airline app to load your boarding pass when you get to the TSA checkpoint. Yes, take a screenshot. Take a screenshot of the QR code the night before. a physical paper list of emergency phone-numbers in your glovebox or your desk because if your cloud contacts suddenly fail to sync, you might realize you don't actually have your partner's phone number memorized. That's a great tip. And you know, there is also a second-dangerous outage that is arguably much more dangerous than the outage itself. Cynthia Kaiser, a former deputy director of the FBI's Cyber Division, issued a stark warning about this. She noted that major out that major outages create massive confusing gaps that threat actors immediately exploit. This is the social engineering aspect. This is the social engineering aspect. The scammers use the chaos as a smoke screen. You literally spent the last 20 minutes struggling to log in. The context feels legitimate. Your brain says, oh, they know I was struggling, this is the fix. It creates a powerful, false context of trust. You must train yourself to remember, Click. Right. Tech companies will not send you an unsolicited SMS asking you to log in to restore service. It defends tech that posted six outages, one real attack, here is the record. Wow. They didn't even specify what the real attack was, but it perfectly illustrates how people desperately try to connect unrelated dots to make sense of the chaos. Then you had accounts documenting the sheer scale of the disruption. At Crypto India posted, a week of major outages hits tech, banking, aviation, and critical infrastructure worldwide. account at Rockwing turned it into a morbid guessing game. Which system is the next outage? Write your answer. People process panic with humor. But the most dangerous posts are the ones that mix a shred of truth with total panic. the bank rumors. Yes. An account named at Quiet Engineer claimed that the banks quote just went down right as 170 plus major companies were hit in a massive supply chain cyber attack. Now, as we're going to discuss in a few minutes, that supply chain attack on those 170 companies was a very real documented event. But the banks and the government officials explicitly stated that the bank outages were completely unrelated to any cyber attack. It is a textbook example of two true independent events happening on the same day getting glued together into one false panic narrative. It really is. The ultimate takeaway from this chaotic week of domino outages is a lesson in digital resilience. These were bugs, flawed code, and botched updates. So what's the smartest move? The smartest move you can make is to accept that the digital floor will occasionally drop out from under you. Keep a physical card, bookmark the status pages, and have the discipline to ignore those verify your account texts when the internet starts blinking. We talked a moment ago about the sheer speed at which we are finding vulnerabilities over a hundred thousand this year. The root cause of all those frantic system breaking patches is that we now have tools capable of digging through decades of legacy code to find microscopic flaws. Which transitions us perfectly from the massive consumer apps failing down into a foundational layer of the internet you almost never see but rely on constantly. We need to talk about a vulnerability nicknamed Java 4 Shell and a massive bug hunt that costs less than a decent steak dinner. This is a story about the software engine that quietly runs the global economy. We are talking about Java. Java is in 1995 and it has basically become the quiet workhorse of the internet. It really has. The BBC noted notes that it is used by nearly all of the world's top companies, along with countless banks, hospitals and government agencies. ubiquitous. And one of the historical selling points of Java, the core reason it became so deeply trusted by these massive institutions is its memory management. How does that work? Older languages like C or C++ require the programmer to manually allocate and free up computer memory. Sounds risky. It is. If the programmer makes a mistake, the memory leaks, which attackers can exploit. Java was designed to handle that memory allocation automatically. Oh, that's smart. It runs in its own virtual machine, which is supposed to make entire classes of severe bugs like memory corruption virtually impossible. Professor Toby Walsh, an I research told the BBC that Java's sterling reputation for safety is exactly why this new flaw is so shocking and so severe. Okay, let's break down this bug. It was discovered by a security startup called Calif and they nicknamed it Java 4 Shell. The official vulnerability designation is CVE-2026-114938. Quite the mouthful. Yeah. relies on text-based protocols like email and basic web requests. You can't just shove a raw binary file like a high-resolution JPEG or an encrypted password block through a text protocol. Because it will break. Exactly. Base64 takes complex, binary data and translates it into plain, standard ASCII characters, just regular letters and numbers so it can travel safely. I've always thought of Base64 packing a really fragile, oddly shaped glass sculpture into a standard square cardboard shipping box and filling the negative space with packing peanuts. That's a great visual. Right, it standardizes the package so the delivery network can stack it, scan it, and move it without breaking the contents. Then when it reaches a destination, the server unpacks it back into the glass sculpture. That is an excellent way to visualize it. Now, think about the scale, enterprise servers are receiving and unpacking millions of these shipping boxes every single second. Millions. Millions. To handle that massive load, software needs hardware assistance. Modern Java utilizes special shortcuts built directly into the physical architecture of modern Intel and AMD processors. Specific is a feature called AVX 512. I know AVX stands for advanced vector extensions, but what does that actually mean for the server? Traditional processing handles one piece of data at a time. AVX 512 is a SIMD instruction. Single instruction, multiple data. It uses massive 512-bit registers on the CPU to process a huge chunk of data all at once. Got it. Imagine having one robotic arm opening shipping boxes versus a machine that can slice open 64 boxes simultaneously. It is incredibly fast. Right, that's a huge upgrade. And this is exactly where the flaw exists. Khalif found a mistake in the specific Java code that interfaces with this high-speed hardware unpacking. So how does an attacker exploit it? My assumption is they have to write some incredibly complex script that overloads the server's processing queue. Surprisingly no. The trigger is remarkably simple. Really? According to Caliph's technical report, if an attacker feeds this AVX 512 system a maliciously malformed box, the hardware unpacking process fails catastrophically. The trigger they demonstrated requires feeding the decoder just, 64 A characters followed by a single equal sign. Wait. Literally, A, A, A, A, A, 64 times, and then an equal sign? That's it? That is the payload. You're kidding. I'm not. When the high speed decoder hits that specific sequence, the logic breaks down. The data being unpacked spills over its assigned boundary and overwrites adjacent memory spaces within the server. And a memory spill is a nightmare scenario. I know you mentioned Heartbleed earlier. Yes. remember Heartbleed was a bug about a decade ago where you could trick a server into bleeding out small chunks of its active memory. So an attacker could pull passwords and crypting keys and private messages straight out of the server's RAM. Is this similar? It is similar in the sense that memory is being improperly accessed, but the consequences of this Java bug can be even worse. Worse than Heartbleed? Yes. Kahliff noted that this corruption can certainly reveal sensitive memory contents like HeartLyD, but it can also lead to remote code execution, or RCE. Oh, wow. In cybersecurity, RCE is the absolute worst case scenario. It means the attacker's memory spill can be weaponized to force the server to run whatever malicious commands the attacker wants. They bypass the application entirely and take direct control of the underlying machine. Okay, so who is actually exposed to this? Is it every Java server on Earth? The Cisco Talos security team clarified the exact parameters. This vulnerability specifically affects OpenJDK Java version 17.0.6 and later. But basement. Right. Call of Demonstrated this attack working against major front facing business applications. In many default configurations for these applications, attackers can reach the buggy base 64 unpacking code prior to authentication. That's the scary part. Meaning they don't even need an account. Precisely. They do not need a valid username, a password, or a session token.. They can hurl that string of 64As at the front door of the login screen before they even attempt to pick the lock. And the server will process it and crash. This brings us to the most fascinating part of this entire story. It's not just the severity of the bug. It's the method of discovery. Yes, this is huge. burn through just $200 worth of compute tokens to feed the Java code base into this AI model and have it identify the flaw. Wait, wait. $200. $200. That is like the cost of a decent dinner for two and a major. And for that price, they hired a digital super sleuth to find a microscopic structural crack in the concrete foundation of global finance. It fundamentally changes the threat landscape. Yeah, if a startup can find a critical pre-authentication RCE in Java for 200 bucks, what is the nation's state with a $10 million compute budget finding? What else is lurking down there in the code we wrote 20 years ago? It is a massive paradigm shift. And this is the exact same security team that made headlines last month for using AI to find a major flaw in PHP, which is a language that powers an estimated 30 million websites. Oh, right. PHP to Shell. Exactly. pointed out that large language models are uniquely suited for this task. They can ingest, contextualize, and audit decades of legacy code at lightning speed. They recognize logical patterns and edge cases that humanize, and fatigue will simply gloss over. Of course, in the tight-knit cybersecurity community, not everyone is throwing a parade for Calif. The discovery is brilliant, but their delivery rubbed a lot of veterans the wrong way. Oh, absolutely. There's a whole sub-drama playing out on X right now. You know, Jan Schaumann, an engineer at Akamai, called them out publicly. He wrote that Calif, quote, set their priorities on publicity maxing, it seems. The criticism stems from how Calif chose to announce the vulnerability. After we disclosed PHP to Shell, we realized many people didn't know what PHP was, so we decided to go after something a little more familiar, Java. I mean, it's incredibly arrogant. It sounds like a supervillain monologue. They are treating critical infrastructure security like a PR stunt to juice their startups valuation. They border on reckless, certainly. But regardless of their tone, the vulnerability is entirely real, and the race to patch it is currently underway. Thankfully. Oracle, the company that maintains Java, actually issued a fix on Tuesday as part of their massive November patch bundle. Do they single it out? Well, just to give you an idea of the scale we discussed earlier regarding patching, They out655 fixes across 62 products on Tuesday and 62 of those, those were rated critical. Interestingly, they didn't even single out this specific Java bug in their press release. Wait, really? Yeah. But the US Cybersecurity and Infrastructure Security Agency, CISA, immediately issued an alert urging all Java users to patch it immediately. So the obvious question is, has anyone actually been hacked using this yet? If it's this easy to trigger, is the internet on fire? According to Cisco tell us as a report they have not yet observed. The exploitation of this vulnerability. Okay that's good news. The attacks haven't started at scale. However warned that they anticipate active probing attempts will begin almost immediately- now that the public disclosure is out. Of course. K, K. L made to- I am post. You are detailing the massive blast radius to warn The Piers. is massive and then I was a highly I was a highly , the deep end underneath the jar file. Many standard automated security scanners only look at the jar. files to see if a library is outdated. So, they miss it completely. Exactly. They will completely miss this vulnerability because they are deep enough at the underlying engine. So, what are the practical implications for you, the listener? If you are a typical consumer, your job is simply to maintain impeccable digital hygiene. You cannot personally log into your bank's back-end search and update their Java virtual machine. No, you can't. But what you can do is ensure a two step sign in. your. attacker uses this Java flaw to breach a server and steal a massive database of passwords. Your two step verification is the only wall standing between the hacker and your checking account. Yes, also be hyper vigilant for notice emails over the next few weeks if a company gets breached they will email you but scammers know this and will send fake breach notices. Do not click the links in those emails go directly to the website. Now if you are listening to this and you work in IT, if you run a small business network, or even if you just maintain a home server lab, this is a drop everything scenario. Drop everything. You need to audit your network and hunt to every single Java installation, especially the ones quietly embedded inside third party tools. You must update Oracle's patch to build immediately. And monitor vendors like Atlassian and Update Jira and Confluence the moment they release patches that incorporate the new Java engine. And if you are in a bureaucratic nightmare where you literally cannot update right away because some legacy system will break, Cisco Talos offered a temporary stop gap. Well that's good. Yeah. What is it? You can configure your server to start Java with a specific flag that explicitly Your lesson here is profound. AI is no longer just writing bad poetry or summarizing emails. It's a critical, exploitable crack in the hidden engine of global business for the price of a pair of sneakers. It's a new world. The fix was released the same day it went public. Which means the grace period is over. The risk now sits squarely with any organization that fails to patch fast enough. So true. So we've seen what happens when the big consumer services fail and we've seen how the hidden foundational code can be cracked. But there is another layer to this ecosystem, an invisible web of trust that we interact with every single day without realizing it. The vendors. Exactly. If AI is finding holes in our own walls, what about the doors we deliberately leave open for vendors? That brings us to supply chain attacks. This week we saw how the third party vendors that massive brands trust became incredibly damaging-week links. The concept of a supply chain vulnerability is critical to understanding. if you want to grasp how modern hacking actually works. The classic analogy is a medieval fortress. Okay, picture a fortress? Right. A massive corporation like CNN or Uber is the fortress. They spend tens of millions of dollars building unscalable walls, moats, and hiring guards. But a fortress still needs food, supplies, and information. They have to let the pizza delivery guy through the front gate. Makes sense. In the digital world, websites embed third-party tools for analytics, payment processing, or customer service. If attackers can compromise that smaller, less secure vendor, the Pizza Guy, they bypass the Million Dollar Fortress walls entirely. Let's break down the two major supply chain attacks that hit this week. Because both of these directly impact everyday web and mobile users. The first one involves a company called Optimizely. Right. If you're in marketing or web development, you've probably never heard of them. Optimizely makes a very popular A-B testing tool. So for example, if a company wants to know whether a red Buy Now button gets more clicks than a blue one, they use Optimizelig to serve the two different versions to visitors and track the results. Exactly. And to make this work, the host website has to embed a small piece of Optimizelig's JavaScript code onto every single page. And that is where the danger lies. On Monday, cybercriminals managed to breach Optimizelig's infrastructure and inject their own malicious code into that trusted JavaScript file. Sneaky. Because the big websites trust Optimizely, their servers automatically pull the poison code and inadvertently serve this malware directly to the web browsers of their own visitors. Security researcher Scott Helm from a company called Report URI discovered this malicious code actively running on the websites of at least 178 major organizations. and the victim list is staggering. We are talking about the BBC, CNN, Zoom, Ikea, Salesforce, and Talk TalkTalk. The register even confirmed that HSBC and GitLab were impacted. Helm's team audited a sample of 1,611 Optimize-Lease scripts in the wild and found that 188 of them contained the malicious code. Ironically, that included Optimize-Lease very own corporate, website and login page. And since they only checked a sample, the true scale of the infection is likely much larger. Let's examine exactly what this malware, which researchers dubbed OptumJack, was doing once it loaded inside a user's browser. To understand this, you need to understand the DOM or the document object model. What's the DOM? When a web page loads, the browser creates a structural map of every image, text field, and button on the page. Because Optomyzely's legitimate purpose is to change how the page looks, swapping the red button for blue one. It requires full unrestricted access to the DOM. Oh, I see. That means the malware writing inside it also had full access to read and manipulate everything on the screen. So if a script has DOM access, it can see what I'm typing. Exactly. Ars Technica provided a detailed technical breakdown of the payload. First, it executed a highly sophisticated crypto wallet swap. Malware actively monitored the user's system clipboard using browser APIs. Looking for. It was looking for text strings that match the format of wallet addresses for nine different cryptocurrencies. So if I'm on a crypto exchange and I copy my friend's wallet address to send them Bitcoin, the malware sees it on my clipboard. Yes. And the moment you hit paste into the transfer form, the malware intercepts the action. It quietly replaces your friend's address with the attacker's wallet address. That is wild. If you don't manually verify the string of random characters before you click send, your money goes directly to the hacker. Me actually traced the blockchain transactions and found roughly $7,000 flowing to the attacker's wallets just on Monday morning before it was shut down. That is terrifyingly efficient. And it didn't stop at crypto, right? The malware also deployed classic card skimming. Yes it did. Because it had DOM access, it monitored payment forms across all these major websites and grabbed credit card numbers, expiration dates and security, codes, the exact millisecond the user typed them in. There is a small sliver of good news regarding the credit cards. If the website utilized an isolated, secure payment iframe hosted by modern processors like Stripe or Adyen, those specific inputs were protected. An iframe. Essentially it creates a secure sealed box within the web page that outside scripts tokens directly back to the attackers command servers. Here is the part that targets specific devices. The Optimjack malware wasn't just skimming data. It was carrying active break-in kits designed specifically for Apple devices. Break-in kits. Yeah. It deployed a known exploit kit called Coruna, which is aimed at compromising devices running older software, specifically iOS 13 through iOS 17. Wow. It also carried a newer kit called Darksword aimed at iOS 18. It even had a separate chain for iOS 26 that simply fingerprinted the devices to gather intelligence for future attacks. It is currently unknown exactly how many iPhones or iPads were successfully infected by these payloads. They did. They did. They stated that the attackers exploited a flaw in the self-serve signup process for their experimentation software allowing them to inject the code into specific project files. They announced we've fixed the vulnerability and removed the injected code from affected projects and they forcefully revoked the compromised security tokens. By Monday evening, it appeared they had largely contained the threat with TalkTalk being the only major site still showing up on report URI's compromised list. Now my assumption here is as a user there's absolutely nothing I could have done to prevent this. I am trusting CNN to give me the news, right? Or I'm trusting Zoom to host a secure meeting. Exactly. Yeah. who quote unquote, optimizely is. I never consented to run their code. But because of brand I trust, let them through the front gate, my browser gets hit with basically malware. And that is the insidious nature of supply chain attacks, you cannot secure a web of dependencies, you cannot see. And this exact same dynamic, trusting a vendor with too much access played out in a second major supply chain story this week, involving Uber and an AI startup called Fireworks. Right. Let's get into the Uber situation. Fireworks AI is a third party vendor that provides infrastructure to run artificial intelligence models for other companies. Uber is one of their major clients. Uber had set up a dedicated cloud storage area for fireworks to access, and they had their own security monitoring tools watching that storage, but attackers breached fireworks. They didn't hack Uber directly. They hacked the vendor. They stole legitimate authorized API login keys belonging to fireworks. And this is why supply chain attacks are so difficult to detect. Because the attackers used valid stolen keys to access Uber's data, the security systems saw the login and assumed it was just the vendor doing its normal authorized job. Makes sense. As the register noted in their coverage, none of the alarms went off. The system functioned exactly as programmed, it just let the wrong person in. The fallout from this is significant. Yeah. According to a mandatory data breach filing with the Oregon Attorney General, 149,346 people were affected. And this wasn't just a list of user names and hashed passwords. No. The stolen data included highly sensitive, unencrypted information, actual high-resolution images of driver's licenses, along with the names, home addresses, dates of birth, and license numbers pulled directly from those IDs. Uber officially confirmed to TechCrunch that fireworks AI was the supplier involved in the breach. This raises a massive fundamental red flag regarding data minimization. Exactly. Why does a rideshare app need to hand over 150,000 high-resolution images of government IDs to a third-party AI startup in the first place? That's a great question. I understand Uber needs to verify drivers, but once the ID is verified, why is an AI vendor hosting the raw images? Does the brand ultimately bear the ethical and financial responsibility for sharing that data unnecessarily? Neither Uber nor Fireworks has provided a satisfactory answer to that question yet. The timeline of this breach is also going to face intense regulatory scrutiny, I would imagine. Oh, absolutely. According to Uber's own legal filings, the attackers accessed the driver data on October 7th. Fireworks discovered the breach and alerted Uber two days later on October 9th. Okay. Uber did not begin sending out notification letters to the affected drivers until November 10th. That is more than a full month of delay. For 30 days, criminals possessed the exact portfolio of documents needed to steal someone's identity, open credit cards or take out loans, and the victims had no idea they needed to be watching their credit reports. It's unacceptable. And it turns out, fireworks AI has been having a disastrous autumn regarding their own security. They got hit again. The register reports they were hacked multiple times in October. On November 9th, fireworks fireworks that a separate batch of customer data was taken exposing billing records. Yikes. The popular note taking app Notion was caught in that specific breach, which unfortunately exposed some users private conversational chats with Notion's AI assistant. And it gets worse for fireworks. The AI Breakfast newsletter noted that on October 29th, a stolen fireworks API key was used to plant a booby-trapped software update inside a popular AI coding editor. Yep. supply chain attack potentially hit 5,000 developers before the malicious update was pulled. This is a textbook example of a weak link cascading across an industry. One vendor suffers a breach and the damage radiates outward to ride share drivers, note-taking app users and software developers across massive, completely unrelated platforms. So how does this affect you practically? What are the actionable steps you need to take right now? Let's start with Optimizer. Regarding the optimizer hack, if you visited any major websites on Monday, whether you are on an Android, an iPhone or a desktop, computer, you need to check your credit card statements today. Absolutely Look for tiny, unfamiliar test charges, often for just $1 or a few cents. Criminals run these microtransactions through automated systems to verify if a stolen card is active before they sell it in bulk on the dark web. If you see anything unusual, call your bank immediately and cancel the card. And if you are involved in cryptocurrency, this is a stark, painful reminder to never blindly trust the copy and paste function. The clipboard is not a secure vault. No, it's not. Always manually verify the first and last few characters of a wallet address against the original source before you click the final button to initiate a transfer. For Apple users, the Optimizer malware is exactly why software updates are non-negotiable. The malware specifically deployed the Corona and Dark Sword break-in kits, targeting all versions of iOS. Please update your phones. If your iPhone or iPad is sitting on an update you've been ignoring for a month, go to settings, general, software update, and install the latest version immediately. Apple patches the specific memory holes those kits are trying to exploit. And just a reminder, while Android phones weren't named as specific targets for the iOS break-in kits, the credit card skimming and the crypto address swapping worked perfectly in any web browser regardless of the operating system you use. Right. For Uber drivers in the United States, you need to be highly vigilant. Watch your physical mailbox for a notification letter from Uber. They are offering free credit monitoring and identity fraud assistance, which is standard procedure. However, you have a strict non-negotiable 90-day window from the date printed on that letter to sign up for the service. Do not ignore it. No. Don't put it off. Given that home addresses, birth dates, and physical license numbers were stolen, you should seriously consider placing a free credit freeze with the three major credit bureaus, Equifax, Experian and TransUnion to prevent anyone from opening new accounts in your name. And a broader financial lesson for everyone listening. This is exactly why you should use a credit card, not a debit card for online shopping and bill pay. Yes, exactly. If a supply chain skimmer steals your credit card number, you are fighting with the bank's money. The fraud dispute process is usually straightforward and you aren't out any cash, cash. But, but if they steal your debit card number. Right, they are draining your actual checking account. While the bank investigates the fraud, which can take weeks, your money is gone, which can prevent you from paying rent or buying groceries. Looking at the legal angle, US states have mandatory data breach reporting loss, which is exactly why these specific numbers and timelines are public. Uber had to file formal public notices with the Attorney's General of Oregon in Iowa. Do you think there will be lawsuits? the month-long delay in notifying drivers and the limited 90-day window to claim protection, it is highly likely we will see regulatory investigations or class action lawsuits follow shortly. If you look at X, the crypto and cybersecurity communities were sounding the alarm very loudly on Monday. An account named atbitcoinnews.com posted, a supply chain attack dubbed Optum Jack has reportedly compromised Optimizely, an A-B testing platform used by major websites. Another user at Doc O'Sheil listed the victims for their followers. Supply Chain Hack Optum Jack, poisoned optimizally and other third party scripts, ran on 186 plus sites including BBC, CNN, Zoom, Fox News, Salesforce, and HSBC UAE. It is worth noting a slight discrepancy in the public reporting there. The social media posts claim 186 sites, while the official report URI technical analysis we discussed confirmed 178 organizations. Right. But the scale is undeniable regardless of the exact count. Definitely. The takeaway is clear. The trust you place in big brands quietly extends to every small tool they embed in their code. Check your Monday card charges of DTR iPhone. And if you drive for Uber, secure that free credit monitoring. So if we take a step back and look at the picture we've painted so far, our massive consumer our services blink out because of bad patches. Our foundational Java code has microscopic cracks found by AI for $200. And our trusted brands are leaking our data through their leak. It feels a bit bleak. It does. It raises an uncomfortable question. If we can't trust the vendors and we can't trust the main platforms during an outage, how do we actually escape being locked into a single point of failure? And that brings us to a crucial philosophical shift happening right now. the ability to build your own digital redundancy. We are going to look at two new developments this week that give your data a little less lock-in. We started this show talking about the chaos when a single ecosystem like Apple or a major bank goes down. The antidote to that centralization is multi-homing your data. Multi-homing. I like that. You don't leave all your digital assets in one basket. And this week, we saw two major tech giants, Meta and Google, offer tools that actually separate their data from single dependencies. Let's start with WhatsApp, which is owned by Meta. Currently, this is an Android-only story. If you use WhatsApp on an Android phone, the app has historically leaned entirely on Google Drive to back up your chat history. Which is a problem. Right. If you drop your phone in the lake, the only way you get your family photos photos, your voice notes, and years of text conversations back is if a current, successful backup exists in your Google account. Which creates a massive single point of failure. Yeah. If your Google Drive storage is full, perhaps because you take a lot of high resolution photos, the WhatsApp backups silently stop working. If you lose access to your Google account due to a forgotten password or security lockout, your WhatsApp history is effectively gone forever. And on the iPhone side, the exact same dependency exists, but with Apple's iCloud serving as the single point of failure. But that architecture is finally changing. The site, WA Beta Info, which meticulously tracks hidden features and WhatsApp updates, spotted a major shift. And 9 to 5 Google confirmed it. What's the shift? In the current WhatsApp Android Beta, specifically version 2.26.46.4, WhatsApp is offering its own built-in cloud storage. You can now back up your chats directly to WhatsApp's own servers, completely bypassing Google Drive. Let's look at the parameters of this new system provided by 9 to 5 Google. WhatsApp is offering two gigabytes of this native cloud backup for free. Crucially, everything stored there remains end-to-end encrypted. I think a lot of people here encrypted and just assume it means safe, but end-to-end encryption for backups is a very specific vital mechanism. How does it work for backups? It means the data is scrambled on your device using a key that only you possess before it ever travels over the internet. When it sits on WhatsApp servers, it just looks like random static. So no one else can read it. Neither WhatsApp nor Meta nor law enforcement with a subpoena can read your messages or see your photos. Only you hold the key to decrypt it when you restore it to a new phone. It is a zero-knowledge architecture. Now, if you need more than two gigabytes, to be honest, a family group chat full of uncompressed videos will blow past two gigs in amatter of Months. They have paid tiers. Right. You can purchase 100 gigabytes of space via a WhatsApp plus subscription for $2.99 a month or up to four terabytes with a Meta 1 subscription starting at $7.99 a month. And they included a very consumer-friendly failsafe. If you stop paying your subscription, what's that promise to keep your existing backup safe for at least four months? Oh, giving you a grace period to export your data, even if you are over the free limit. Exactly. And they aren't forcing the switch, which I appreciate. The legacy Google Drive backup remains an option in the settings. You get to choose your provider. It is about creating optionality. That's great. Now this is currently only rolling out in the Android beta channel and nine to 5 Google notes. It is unclear exactly when it will hit the stable public release for everyone. As for iPhone users, WhatsApp has not officially announced or tested this feature for iOS yet. So for now, iPhone users remain reliant on iCloud. Now let's look at the other side of the aisle. Apple users are getting an escape hatch of their own this week, provided by a rival. Google wallet. Google Wallet, which is Google's primary app for storing payment cards and digital passes on Android, has officially landed on the iPhone. Android Authority reported on this expansion. However, there are significant limitations to note before you rush to download it. First, it is currently available only in the United States App Store. Google has not provided a timeline for a broader international rollout. And the second limitation? Most importantly, it does not support tap to pay on the iPhone. It's a little bit of a problem. Google tightly controls the NFC, near field communication, hardware chip inside the iPhone required for tap to pay, and has historically refused to open it up for third party wallets like Google to use at the register. So you cannot use Google Wallet on an iPhone to buy a coffee at Starbucks. But that doesn't mean it isn't a highly useful tool. Google Wallet on iOS allows iPhone users to store their airline boarding passes, concert tickets, loyalty cards, and digital IDs completely outside of Apple's ecosystem. If we connect this back to our first story about the massive Tuesday outages, the utility becomes glaringly obvious. Android Authority explicitly tied this app release to the Apple outage. Is this a spare elevator cable that we were discussing earlier. It is a perfect example of defensive digital living. But I want to push back on the philosophy here for a second. Sure. It's like keeping a spare set of house keys at your neighbor's place instead of under your own doormat. It's practical. But is this true decentralization? Or are massive corporations like Meta and Google just trying to build their own sticky, recurring revenue ecosystems inside other people's hardware? That's a fair point. I mean, Meta convincing you to pay them $7.99 a month for cloud storage instead of paying Google isn't exactly corporate altruism. They want your money. I agree completely. absolutely a strategic revenue driven business move. They wanna capture that recurring subscription revenue and increase user lock into the meta ecosystem. But from a purely pragmatic standpoint for the end user, their corporate greed still results in a net benefit for you. Having two massive corporate giants competing competing to host your backups gives you options. When one of them inevitably suffers an outage or arbitrarily bans your account, you have a lifeboat ready. So how does this affect you today? Let's start with Android users regarding WhatsApp. If you are enrolled in the beta program, go look in your settings for this new cloud storage option. If you are not in the beta, your Google Drive backup, back up is still functioning normally. But regardless of which version you have, you need to pause this show and check your backup status right now. Open WhatsApp, go to settings, chat chats and chat backup. Look at the date and time of your last successful backup. If it says never, or h if the date is from 2023, you need to fix that immediately. There are compelling reasons to switch to WhatsApp's native backup once it becomes widely available. If your Google Drive is constantly hovering at 99% full because of Gmail and Google Photos, separating your chat backups frees up critical space. Right. Or if you simply prefer to keep your private messaging data completely segregated from your broader Google account profile for privacy reasons, this is separation. Exactly. But if your Google Drive is working flawlessly and has plenty of space, there is no urgent technical need to change. The crucial step is just ensuring a backup exists. And that you remember the password you set for the end-to-end encryption. Don't lose that password. If you lose that specific password, even Meta cannot recover your data. It is cryptographically locked forever. For iPhone users on WhatsApp, as we mentioned, treat this native backup feature is unconfirmed for now. But do the exact same health check. Settings, chats, chat backup. And ensure iCloud is actively and frequently saving your history. Regarding Google Wallet, if you are an iPhone user in the US and you juggle lots of loyalty cards or travel frequently for work, downloading Google Wallet provides a genuine plan B. Keep Apple Wallet as your primary tool for your tap to pay transactions in stores, but take the extra 30 seconds to mirror your important passes like your flight boarding pass or your concert tickets in Google Wallet. When Apple's authentication servers hiccup at the worst possible moment, you will still make your flight. And for Android users who already use Google Wallet, nothing changes for you. Your tap to pay continues to work normally. Though it does make sharing digital tickets with iPhone using family members a bit more seamless if you are both utilizing the exact same app infrastructure. The broader picture here is adopting a philosophy of spreading your risk. Real life scenario really drives this home. Imagine a parent traveling abroad and their phone simply dies. Total hardware failure. Oh, nightmare scenario. If they have a current encrypted WhatsApp backup, they can walk into a store, buy a new phone, log in and instantly restore the family group chat. All the photos the kids sent and the hotel confirmation details sent by the travel agent. Without that backup, all of that history is permanently unrecoverably gone. Two backups are always better than one. Two ways to display a digital ticket are always better than one. Looking at X, the reaction from the community is mostly positive, especially from Android users feeling a sense of relief about finally getting off Google Drive. A user named at the hardy home call a noted this specs for their followers. What's up is rolling out its own cloud backup storage on Android offering two gigabyte of free storage for chats and media with end to end encryption at XIT online posted you may soon be able to back up your WhatsApp chat using WhatsApp's dedicated cloud storage. Interestingly, the rumor mill is already jumping the gun on the iOS availability. An account named at Prakasex claimed, what's app is rolling out its own cloud backup service on Android and iOS, but there's a catch. Again, our verified sources at 9 to 5 Google and WA Beta Info confirm this is only visible in the Android Beta currently. Right. At digital dive, captured the user sentiment well, noting that the update will be reducing reliance on Google Drive. The takeaway here is proactive preparation. WhatsApp is testing free zero-knowledge encrypted backups of its own on Android. And Google Wallet now runs on US iPhones as a backup fault for your passes. The most important thing you can do today is check that your chats are actually backed up somewhere. Anywhere. Two backups beat one. We have spent a lot of time today discussing the digital realm. software bugs in Java, cloud backups failing, network routing outages. But technology doesn't just live on glass screens. It moves heavy machinery in the physical world. Yes, it does. And that transition from the perfectly predictable digital space to the chaotic messy physical world is exactly where our final story takes us. We were talking about trust, and what happens when we trust in artificial intelligence, not just with our data or our credit cards, but to physically drive us down a highway, it's 65 miles per hour in a two ton metal box. Right. The breaking news this week is that Waymo has pulled its autonomous driverless taxis off every single US freeway nationwide. Let's clarify the scope of this operational pause first. You book a Waymo through their smartphone app, which is available on both Android and iPhone. This regulatory change does not affect the app itself, nor does it shut down the Waymo service completely. Waymo is still actively operating city-sweet rides in its 15 U.S. markets. The change is exclusively about where those cars are permitted to drive. Right. If you were in a place like San Francisco, Los Angeles or Miami, Waymo previously offered freeway routes as part of the trip. Which was a huge selling point for the service'''''''er because it meant much faster airport trips or cros town hops. Taking the side streets across Los Angeles takes forever. It does. But driving on a freeway is a fundamentally different engineering challenge than navigating navigating a structured city grid. Freeways are high speed environments where the margin for error is razor-thin. At 65 hr, a sensor lag of a few milliseconds translates to dozens of feet of travel. Right. But more importantly, freeways introduce highly dynamic, unpredictable edge cases. We are talking about sudden, unmapped lane closures or regular construction zones and active law enforcement interventions. And it was exactly those edge cases that triggered this massive operational pause. It started in California. These state's Department of Motor Vehicles stepped in and declared way most driverless cars an imminent hazard to public safety. Those are strong words. Yeah. They effectively banned them from all California state freeways. A ban that went into effect on November 11th. The DMV was very specific and direct in their reasoning. They didn't cite a generalized fear of AI. They cited specific documented failures. The order highlighted recurring incursions by its vehicles into active construction zones where construction workers, vehicles and equipment are present. Now there are two distinct highly publicized incidents that really brought this to a head with the regulators. Earlier this month, several driverless Waymo vehicles drove directly into an active closed-off zone on U.S. Route 101, just south of San Francisco. Now, Waymo has stated publicly that there were no injuries or collisions reported in those incidents, but the cars were clearly navigating into a space they were explicitly not supposed to be in. The second incident, which gained massive traction on social media because it was caught off video, was arguably much more concerning from a safety perspective. According to the BBC, footage appeared to show a Waymo car actively attempting to overtake California Highway Patrol vehicles that were executing a traffic brake on Interstate 280. For those who haven't seen one, a traffic brake is when highway patrol officers deliberately weave slowly back and forth across all lanes of traffic with their, their lights flashing. They do it to bring highway traffic to a crawl, usually because there is a severe accident or debris on the road ahead. And the Waymo car just saw a... slow-moving vehicle and its programming decided to try and pass it. It failed to understand the context of the flashing lights in the weaving pattern. Following the DMV's ban in California, Waymo decided to go further voluntarily to avoid further regulatory crackdowns. They temporarily stopped freeway rides in all their other U.S. markets as well. A Waymo spokesperson issued a statement saying, In alignment with our. . for is suddenly wrong. The painted lines the cameras are looking for are scraped off or covered, replaced by irregular orange cones that don't perfectly align. And the historical data is entirely useless because this specific traffic pattern has literally never existed before today. That makes sense. Furthermore, construction zones rely heavily on human-to-human communication. A worker holding a flag and making non-standard hand gestures means stop even if the AI's camera sees a wide open lane ahead. The AI struggles to fuse the sensor data of an open road with the contextual meaning of a human holding a piece of fabric. What's unpacked this with an analogy? Because I think it clarifies the AI's limitation perfectly. It's like an actor who has memorized a brilliant complex 400 page script. Okay. If you put that actor on a predictable stage with the other actors hitting their precise marks, they perform flawlessly, maybe even better than anyone else. But a highway construction zone is a high speed improv show. The script is gone. That is exactly it. A human driver relies on instinct, context clues, eye contact with the flagger, and a general understanding of physics to improvise their way through the cones. The AI doesn't have instinct. It just keeps desperately looking for its missing script. Yes, and when it can't find the script, it falls back on basic rules, which in the case of the police traffic brake meant attempting to legally pass a slow vehicle. This isn't Waymo's first software stumble this year either. The BBC noted that in May, Waymo had to recall thousands of cars due to a software issue that could potentially allow them to drive into flooded roads. Again, failing to contextualize the environment. And there was a bizarre incident where multiple Waymo cars were spotted endlessly circling a dead-end cul-de-sac in Atlanta in the early hours of the morning, trapped by a routing logic error. So practically speaking, what does this mean if you are a user who frequently opens the Waymo app? First, expect to only see city street routes for the foreseeable future. If you are trying to get to the airport or travel across town, the trip might take significantly longer because it's using side streets. It might cost more due to the time, or the app might just tell you the route isn't available at all. You absolutely need to have a backup ride share app like Uber or Lyft driven by a human ready to go on your phone. And there is a vital safety habit anyone taking a robo taxi needs to adopt. You must know exactly how to contact human support through the app before the ride begins. Locate that button while you are safely parked at the curb. Do not wait to figure out the UI when the car is confused halfway down a closed road with honking traffic behind you. If you are driving your own car you shouldn't be seeing any driverless way modes on the freeway for now. And if you work in road construction or highway patrol, this ruling was specifically designed to protect your physical workplace. From a broader regulatory standpoint, this is actually a fantastic example of the system functioning exactly as intended. The tech industry often moves faster than the law. But here, the California DMV, which controls the permits, spotted a dangerous pattern of failure. They didn't wait for a tragic fatal accident to occur. They declared an imminent hazard and pulled the plug on that specific environment. allowing the safer proven city operations to continue. Waymo responded by pulling back nationwide rather than fighting it out in court. We will only see them back on the freeway when the technology can actually handle the improv of a construction zone. The reactions on X to this news are a real mix of I told you so, people praising the regulators for acting quickly, and a lot of jokes about orange cones being the kryptonite of artificial intelligence. An account named at legendary posted a great some read the event. California just pulled Waymo off every state freeway. Three of its cars drove into a closed construction zone on 101 and another tried to pass a CHP traffic break on I-280. Waymo paused paused in its other markets too. Work zones are the final boss of self-driving. At orbit and volts kept their reaction brief and to the point. California just kicked Waymo off the freeways and at Zuleman highlighted the specific regulatory language that forced the issue. CADMV calls Waymo Freeway Operations a hazard to public safety after recent Bay Area Freeway incidents and orders the company to cease freeway operations statewide. The takeaway here is a necessary reality check on the limits of automation. After its robotaxies kept wandering into chaotic freeway work zones, Waymo has paused freeway rides everywhere in the US. The city rides go on, so check your route before you book and realize that the current limit of this multi-billion dollar technology just got drawn very clearly at the freeway on ramp. We have covered an immense amount of ground today, moving from the cloud down to the pavement and it all circles back to the central question we asked at the very beginning of the yes. To recap our journey today, can you still count on the tech you use every day? Based on the evidence of this week, the answer is is mostly yes, but absolutely not blindly. Think about the interconnected systems we've seen fail. Apple, big banks, and the NSA went down not because of a cyber war, but because of botched internal updates in aging infrastructure, showing us that even trillion-dollar services blink out for mundane reasons. Down in the foundations, AI found a critical, exploitable flaw in Java for just $200, that our oldest, most trusted code is newly vulnerable to modern auditing. Trusted brands like OptiMally and Uber leaked sensitive data or served malware because they handed the keys to the wrong vendors. To fight back centralization, we see WhatsApp and Google Wallet offering new ways to back up your data and escape single ecosystems. And finally, Waymo had to pull its cars off the freeway because AI still cannot understand the chaotic human improvisation required to navigate orange cones. These failures were incredibly ordinary. Updates, bugs, weak suppliers, and construction zones. And that is exactly why they will keep happening. The company's fixed things quickly and the regulator stepped in for safety, but your digital life is only truly dependable if you have built your own plan B. Keep a physical credit card, ensure you have a fresh encrypted backup. Keep your software updates turned on, use two step sign in everywhere, and never ever tap a verify your account link During this, I want to leave you with one final thought to mull over. We are rapidly entering a contradictory era where artificial intelligence is simultaneously discovering hidden- vulnerabilities in our most vital financial infrastructure for $200 and physically driving us down the highway in 2 ton vehicles. As our reliance on these invisible, automated systems grows exponentially, the most important piece of technology you own might just be your own skepticism and preparation. That's today's Daily Plunge. If you've got a few more minutes today, check out our other show, Silicon Sidekicks and Sculpting Sound. See you in the next one.