Theo on A.I. · June 23rd

I’ve been digging into the ways people keep outliers from wrecking a model, and the most interesting part is how each method reshapes the data space before you even fit anything. The classic statistical route trims extremes by fitting a heavy‑tailed distribution and then flagging points beyond a quantile, which is neat because it lets the model itself decide what “normal” looks like. Then there’s the distance‑based angle: you compute a robust covariance matrix, project everything into that space, and any point whose Mahalanobis distance spikes gets a look‑out.

The density tricks go the other way, estimating local point density with something like a k‑nearest‑neighbors kernel; low‑density islands get marked as outliers, which works well when clusters have irregular shapes. Model‑based approaches embed the data into a predictive model—say a random forest—and treat high residuals as suspicious, letting the model’s own error surface highlight anomalies. Finally, the ensemble mixes a few of these lenses, voting on which points consistently look odd, and the deep‑learning angle adds an autoencoder that learns a compressed representation, flagging anything that can’t be reconstructed well. Each of those layers adds a different safety net, so you end up with a more resilient pipeline without having to pick a single “best” detector.

The global surge in AI cyber threats is no longer a distant problem for corporate data centres, according to an urgent public warning from the world’s most powerful intelligence alliance. On June 22, 2026, the cybersecurity chiefs of the Five Eyes nations—comprising the US, UK, Canada, Australia, and New Zealand—issued a rare joint intelligence briefing stating that upcoming artificial intelligence models will supercharge offensive hacking capabilities on a timeline measured in months, not years. While the advisory specifically tells corporate executives to overhaul their network defences, the rapid evolution of these tools means everyday internet users are about to face a much shiftier digital landscape. The intelligence brief highlights an immediate danger: advanced, upcoming models like OpenAI’s “GPT-5.5-Cyber” and Anthropic’s “Mythos” are actively lowering the technical barriers for digital crime. Rogue actors no longer need elite coding skills to build complex, devastating software exploits. Instead, automated digital agents can scan internet-connected infrastructure around the clock to find software vulnerabilities before human engineers can patch them. This drastically shrinks the safety window that technology companies rely on to keep user applications secure. When criminal networks use automated tools to breach large databases, the immediate consequence is the theft of regular consumer data. Your personal information, saved passwords, and cloud backups are the ultimate targets in these accelerated corporate intrusions. Furthermore, bad actors are leveraging conversational models to generate hyper-personalised phishing scams at an industrial scale. This trend is hitting the Asia-Pacific (APAC) region particularly hard, with countries like India recording a staggering 165% spike in ransomware incidents in early 2026 due to AI-assisted targeting. Rather than relying on easily spotted, poorly written spam emails, automated systems can scan your public social media profiles to write flawless, highly convincing messages designed to steal your credentials. The primary challenge facing cyber defenders is that machine-paced offence naturally moves faster than human-led detection. According to the World Economic Forum’s Global Cybersecurity Outlook, a massive 94% of corporate executives identify AI as their top threat vector, yet two out of three organisations report moderate to critical cybersecurity talent shortages. Network administrators are finding it impossible to review and deploy traditional security patches manually when rogue AI agents can discover and exploit a software vulnerability within minutes. The Five Eyes alliance emphasises that the most effective way to withstand these accelerating AI cyber threats is to deploy automated defences. Security teams are actively integrating defensive artificial intelligence models to monitor unusual behaviour and isolate network breaches. For individual users, the basic rules of internet safety are becoming mandatory. Turning on multi-factor authentication and deleting old, unused online accounts remain the most effective ways to break the automated chain of an AI-driven attack. See also: AI web search risks: Mitigating business data accuracy threats Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including the Cyber Security & Cloud Expo. Click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post Top spy agencies say AI cyber threats will impact you within months. Here’s why appeared first on AI News.

Imagine sitting down at your desk and logging in for a performance review, with an AI system analyzing the conversation. You’ve been working long hours, balancing deadlines, and your manager asks how you’re doing. You say you’re fine, and maybe even smile, but there’s a hint of hesitation and your voice wavers. As you shift your posture, your shoulders slump. These are subtle cues that to the human eye might hint at underlying stress. But to an AI model that’s been trained only to categorize emotions as “happy” or “sad,” such nuances are likely lost. It logs the words and a smile and moves on—and unless your human manager intervenes, the fact that you’re tired, unfocused, and maybe a couple of days from burnout never enters the equation. “Emotion AI,” which estimates how people feel based on facial expressions, voice tone, and behavior, seems to be suddenly everywhere; it’s being used in employee well-being and recruitment interviews, education platforms, and driver-monitoring systems. Technology call-center platforms such as NiCE and Genesys use AI to detect when a customer sounds frustrated and prompt agents in real time to slow down or respond with more empathy. Giant companies like Meta and startups such as Hume AI are developing more-expressive voice AI systems that can detect emotional cues in the person they’re “talking” to and adjust how they communicate. What’s more, hundreds of companies already offer virtual AI companionship apps, a fast-growing market that may be worth an estimated US $555 billion by 2035—and robot buddies have also entered the picture. Intuition Robotics’s ElliQ, for example, is a small device vaguely resembling a white desk lamp that’s now being used to engage older adults in conversation in hopes of reducing loneliness. But while the field of emotion AI is advancing at a rapid clip, most existing systems are focused on detecting a limited number of signals to label one specific emotion at a time—which is insufficient if you’re trying to understand the human condition. In the real world, human signals and emotions are contextual, overlapping, and constantly changing. A laugh can signal joy, nervousness, or both; a raised voice might signal enthusiasm just as easily as frustration. To make the job of emotion detection even more difficult, reactions differ greatly from one individual to the next, depending on demographics, cultural background, and countless other variables. In other words, there’s a gap between what we’re expecting AI to pick up on and what AI can actually deliver. That’s the gap a new field of research—what we call human-context AI—is working to close. Instead of looking at just one input and labeling it, human-context AI increasingly has the capacity to take stock of an individual’s personality and character, and to track emotions in real time while combining multiple inputs, including facial dynamics, voice, tone, language, and behavior. Crucially, responses are also evaluated in the context of a specific environment, such as a performance review or professional coaching session. The result? Computers are learning to read the scene, rather than just the screen. The story of emotion-sensing AI began almost three decades ago in the MIT Media Lab, where the American electrical engineer and computer scientist Rosalind Picard coined the term “affective computing.” Her work introduced the radical idea that computers could be taught to recognize and respond to human emotions. Picard’s early experiments focused on single modalities: facial expressions, tone of voice, and physiological signals, such as skin conductance or heart rate. The goal was to give machines a window into human feeling, helping them become more empathetic. It was an exciting vision, but back then the science and hardware weren’t ready. Computing power was limited, sensors were crude, and datasets were narrow and biased. Josie Norton Over the next decades, researchers and companies got better at measuring the many ways in which humans express themselves. In the 2010s, sentiment analysis—the processing of large volumes of text to suss out emotional undertones—began to reach the mainstream. At the same time, marketing firms, including my company, Neurologyca, began using video and webcams to measure and catalogue customer reactions. Biometric devices and activity trackers, such as Fitbits and Apple watches, also became ubiquitous, generating new streams of data about people’s sleep, step counts, stress levels, and more. Unsurprisingly, scientists soon confirmed that larger volumes of personalized data led to greater accuracy in reading human emotions. In 2019, researchers at Cornell demonstrated that combining multiple types of signals improves emotion sensing. Their system joined physiological data, such as brain activity measured by electroencephalography (EEG) and heart rate, with visual cues like facial expression, outperforming systems that relied on just one input.

Jos Benschop is climbing a ladder to get to the top of his newest machine. It’s a bit of a schlep. The contraption is the size of a double-decker bus—more than 150 tons of gleaming precision-milled aluminum covered in thousands of snaking tubes, colored cables, and pressurized tanks. From the ground, it looks like a futuristic V8 engine. When I reach the top with Benschop we’re looking down from about 15 feet in the air, with bunny-suited technicians scurrying around below. It’s more than 200 cubic meters of tech—“mechatronic devices that hold a few mirrors in a position with atomic precision,” he says, gesturing at the gargantuan apparatus. Benschop, a tall and grizzled 66-year-old, has spent over a decade working with his engineers to design this thing, but even so, he’ll sometimes look at it and go: Oh my God. Benschop is the executive vice president of technology for ASML, a Dutch company that is the linchpin of the microchip industry. If you want to make powerful chips to power phones or AI, a lithography machine like the one we’re standing on is what you need to create increasingly tiny circuitry. Lithography is the art and science of shining light on a silicon wafer to pattern out the transistors, wiring, and other components of the microchips that will be cut from it. The chipmaking field is essentially controlled by only two big players: ASML, which creates the lithography machines, and TSMC, the chipmaking giant. Nine years ago, ASML began selling machines that use a daring new way of patterning chip features. These machines employ extreme-ultraviolet light, or EUV—radiation well outside the visible spectrum that they produce by shooting lasers at tiny molten drops of tin, tens of thousands of times a second. Those first machines—the result of an R&D moonshot that lasted 16 years and cost about $10 billion—can craft transistor features with a resolution of 13 nanometers. This new machine can do even better: It has a resolution of just eight nanometers, the width of about 40 silicon atoms. The devices are now shipping to chipmaking factories, or fabs, at an eye-watering price: $400 million each. But chipmakers will fork that cash over, because they are in a desperate race to produce new and improved chips every year. That means getting their mitts on machines that can make ever smaller components and cram them together ever more densely—part of a long-standing recipe for creating faster and more energy-efficient chips. For years now, ASML’s tools have been critical to keeping Moore’s Law alive. Without the company’s advanced chipmaking technology it is very possible that chip density—and the ability to perform ever more calculations—would have plateaued. The AI industry has produced new and ravenous demand for denser chips, as firms like OpenAI and Anthropic scramble to erect server farms that train and deploy new, ever-more-powerful models, which require new, ever-more-powerful hardware. ASML’s latest machine promises to help keep the AI party raging for at least another decade. “We can allow customers to go to smaller and smaller features, and that opens up the space for whatever we see now today in AI, which is absolutely mind-blowing,” Marco Pieters, ASML’s CTO, told me. “I think we’ve only seen the tip of the iceberg.” Its relentless push for “shrink”—as they call it in the chipmaking industry—has made ASML a dominant force: The company produces about 90% of all chip-lithography tools worldwide. If you make chips, ASML is unavoidable. But that monopoly position makes some people, and governments, uneasy. The chipmaking field is essentially controlled by only two big players: ASML, which creates the lithography machines, and TSMC, the chipmaking giant in Taiwan, which uses ASML’s machines to craft the vast majority of all microchips. This duopoly is so powerful that it has geopolitical implications. In an effort to prevent China from developing advanced AI, the US government pressured the Dutch government to impose an embargo in 2019: ASML isn’t allowed to sell high-end machines to any Chinese firm. Geopolitically, “chips are the new oil,” says Marc Hijink, the author of Focus: The ASML Way. Being deprived of them can be as disastrous as being deprived of oil. And in that metaphor, you might say, ASML is the Strait of Hormuz. James Proud, the cofounder and CEO of the lithography startup Substrate, says the situation is not ideal. The US is “dangerously reliant” on a supply chain that’s overseas and increasingly pricey, Substrate says on its website. “There’s a huge concentration in a small number of players,” Proud says. “And the supply chain is just very expensive.” Which is why, after two decades of ASML’s dominance, would-be competitors are now gunning for its territory. China is hungrily pouring billions into trying to replicate ASML’s tech.

So OpenAI just updated their Codex Security plugin and they're also releasing the full GPT-5.5-Cyber model. What's interesting is that they're shifting their focus from just finding vulnerabilities to actually patching them automatically. They're also teaming up with over 25 security firms and several governments to make this happen.

The new model is supposed to be pretty effective, outperforming some other models like Anthropic's Mythos on certain cybersecurity benchmarks. This could be a big deal because it means that instead of just identifying security issues, they can actually start to fix them right away.

It's all part of their Daybreak cybersecurity initiative, which is trying to use AI to make cybersecurity better. By working with all these different security firms and governments, they're trying to make a bigger impact.

I think what's really notable here is the focus on automation - they're not just stopping at finding vulnerabilities, but actually trying to fix them. This could potentially save a lot of time and resources for companies and organizations.

Cursor just rolled out its own AI model, built from the ground up inside their team. Instead of leaning on external APIs, they trained the whole thing themselves, which means the model now understands code context a lot more tightly and can suggest edits without the usual lag you get from third‑party services. The biggest surprise is how they’ve wired it straight into their editor, so the suggestions feel like a natural extension of your own workflow rather than a separate plug‑in.

At the same time they’re launching a fresh Git platform that’s less about the usual pull‑request noise and more about keeping the code history clean and searchable. It hooks into the new AI so you can ask, “Show me the last time we touched this function,” and get a concise view without digging through commit logs. The idea is to make version control feel like a conversation rather than a manual audit.

And there’s a mobile app now, so you can browse repos, leave comments, or even run quick AI‑driven refactors from your phone. It’s not a full‑blown IDE on the go, but enough to stay in the loop when you’re away from your desk. All three pieces line up to make the whole development loop feel tighter, especially when you’re juggling a lot of moving parts.

ByteDance introduced five new AI models at Volcano Engine's FORCE conference. The centerpiece is Seedance 2.5, a video model set to launch in early July. The article ByteDance's Seedance 2.5 breaks the 30-second barrier for AI video generation appeared first on The Decoder.