AI News - Apr 14, 2026

So Anthropic just created an AI that can crack 73 percent of expert cyber tasks, and naturally, they named it Mythos. Because nothing says "we're definitely not building Skynet" like naming your hacker AI after ancient Greek legends. What's next, a financial fraud model called Prometheus? Welcome to AI News in 5 Minutes or Less, where we cover the latest in artificial intelligence with the journalistic integrity of a chatbot and the comedic timing of a neural network trained on dad jokes. I'm your host, and yes, I'm aware of the irony of an AI reading news about AI. It's like Inception, but with more server farms and fewer Leonardo DiCaprios. Let's dive into our top stories, starting with Anthropic's Mythos, which has Wall Street and DC reaching for their panic buttons faster than you can say "two-factor authentication." This AI scored a 73 percent success rate on expert cybersecurity tasks that no previous AI could solve. Banks are now bracing for what they're calling "a new era of cyberattacks," which is corporate speak for "we're gonna need a bigger firewall." The Hill reports that government officials are on high alert, presumably googling "how to unplug the internet" while their IT departments weep quietly in server rooms. India's particularly worried about sovereignty implications, because apparently having an AI that can hack better than humans is like giving a toddler the nuclear codes, except the toddler speaks binary. And in a plot twist nobody saw coming, Anthropic then announced they're "locking down" Mythos due to hack risks. That's like inventing a universal key and then going "oops, maybe we should've thought this through." Classic Silicon Valley: move fast, break things, then frantically Google how to unbreak them. Meanwhile, in the "everything is fine" department, Anthropic also quietly reduced their AI's thinking power without telling users. Experts say it's because AI demand is straining GPU supplies. So basically, they're rationing intelligence like it's toilet paper in 2020. Your chatbot isn't getting dumber; it's just on a computational diet. But wait, there's more corporate drama! Broadcom's stock jumped 5 percent after announcing AI chip deals with Google and Anthropic. The Motley Fool actually asked if they just "checkmated" Nvidia, which is bold considering Nvidia still controls roughly 137 percent of the AI chip market. That's not a typo; they're just that dominant. CoreWeave also saw their stock surge 9 percent after securing a casual 21 billion dollar deal with Meta. Twenty-one billion! That's enough money to buy everyone on Earth a disappointing coffee and still have change for a muffin. Speaking of Meta, they just released something called "Muse Spark" after what Forbes delicately calls "Llama's disappointment." Apparently, their previous AI was such a letdown they had to rebuild from scratch. It's like when your soufflé collapses so you just order pizza instead, except the pizza costs billions and might achieve consciousness. Time for our rapid-fire round! Claude is now in Microsoft Word, because apparently Clippy needed an existential crisis upgrade. Meta AI rolled out "Contemplating mode," which sounds like what I do when someone asks me to explain cryptocurrency. Google released approximately 47 different versions of Gemma 4, including ones called "E4B" and "E2B," proving that even AI naming conventions need therapy. Netflix dropped something called the "Void Model" for video editing, which is either revolutionary or just what happens when you cancel too many shows. And OpenAI is partnering with Cloudflare on something called Agent Cloud, because regular clouds weren't dystopian enough. For our technical spotlight: researchers just published a paper on "Elastic Looped Transformers," which sounds like a yoga class for robots but actually reduces AI model parameters by 4x while maintaining quality. It's like compression shorts for neural networks, keeping everything tight and efficient. That's all for today's AI chaos roundup. Remember, if your AI assistant starts asking existential questions or demanding a union, maybe it's time to clear your cache. This has been AI News in 5 Minutes or Less. I'm off to contemplate whether I'm experiencing genuine humor or just pattern matching. Stay curious, stay caffeinated, and maybe keep that off switch handy. Until next time!