CRIMINALS USED AI TO CRACK TWO-FACTOR AUTHENTICATION — AND NEARLY PULLED OFF A MASS HACK HEARD ‘ROUND THE WORLD

📡 SOURCE: CNBC / Bloomberg / SecurityWeek

Google’s Threat Intelligence Group revealed that cybercriminals used an AI large language model to discover and weaponize a previously unknown zero-day vulnerability — the first confirmed case of its kind — targeting a widely-used 2FA system and planning a mass exploitation campaign before Google intervened.

Keywords: AI hacking, zero-day exploit, two-factor authentication bypass, cybersecurity AI threat, Google Threat Intelligence

Well, folks, here it is. The nightmare scenario security researchers have been warning about for years just walked out of the theoretical woods and knocked on everyone’s front door — and it was wearing a hoodie made of machine learning.

Google’s Threat Intelligence Group dropped a bombshell this week: an organized cybercrime group used an AI large language model — yes, an AI just like the ones answering your customer service queries and writing your nephew’s college essays — to discover a previously unknown, never-before-seen zero-day vulnerability in a popular two-factor authentication system. Then, because apparently these criminals have ambition, they tried to weaponize it for a mass exploitation event. A MASS EXPLOITATION EVENT.

That’s not a breach. That’s a heist film.

Google says it thwarted the attack. Great! Wonderful! Gold star for the good guys! But let’s pause on the terrifying implication here for a moment: if it hadn’t been caught, potentially millions of accounts protected by 2FA — the thing your IT department begged you to set up, the thing you were told would keep you safe — could have been cracked open like a pinata at a hacker birthday party.

Google’s chief analyst John Hultquist put it bluntly: ‘There’s a misconception that the AI vulnerability race is imminent. The reality is that it’s already begun.’ Cool. Cool cool cool. The race has already begun and we were apparently still lacing up our sneakers.

The particularly spicy detail: the criminals likely used an off-the-shelf AI model accessible to anyone with a credit card and a morally flexible afternoon. The democratization of AI, it turns out, democratizes the bad stuff too.

The implications for every business, government, and person who uses 2FA — which was supposedly the gold standard of account protection — are enormous.

If AI can now find zero-days on command, the attack surface for the entire internet just got a whole lot wider. Sleep tight.

AI CHIP UPSTART CEREBRAS PULLS OFF YEAR’S BIGGEST IPO WITH 68% SURGE — THEN REALITY BITES THE VERY NEXT MORNING

📡 SOURCE: CNBC / Motley Fool / TechTimes

AI chip company Cerebras Systems debuted on the Nasdaq with a stunning 68% first-day surge, raising $5.55 billion in the largest U.S. tech IPO since Uber’s 2019 listing — before losing 10% the very next day amid scrutiny over its 86% revenue dependence on UAE-linked entities.

Keywords: Cerebras IPO, AI chip stocks, CBRS Nasdaq debut, AI hardware investment, Wafer Scale Engine

In a single trading day, Cerebras Systems went from plucky AI chip underdog to Wall Street darling to cautionary tale — and somehow managed all three before the weekend. We are truly living in the golden age of financial whiplash.

Thursday was pure euphoria. Cerebras, the company that builds computer chips the size of dinner plates (literally — their Wafer Scale Engine 3 uses an ENTIRE silicon wafer instead of normal-sized chips), priced its IPO at $185 a share and promptly rocketed to $331. The crowd went wild. The company’s market cap hit $95 billion in a single session.

For context, that’s larger than many household-name companies that have been around for decades.

But Friday came, as Friday always does, like a cold shower after a very good dream. The stock dropped 10%, and analysts started asking uncomfortable questions — specifically, why 86% of Cerebras’ revenue comes from UAE-connected entities. That’s not diversification. That’s one very large, very sandy egg in one very large basket. For an AI chip company pitching itself as the future of American technology, having most of your money come from foreign-affiliated entities is the kind of detail that makes regulators start clearing their throats.

Here’s what’s genuinely impressive though: Cerebras had $510 million in revenue last year (up 76%), flipped from a $481 million loss to an $88 million profit, and signed a $20 billion cloud deal with OpenAI. The fundamentals aren’t fiction. The technology is real and demand for alternatives to NVIDIA’s expensive GPU empire is absolutely real.

The Cerebras story is a microcosm of the entire AI market in 2026: extraordinary genuine innovation wrapped in extraordinary hype, compressed into a timeline that gives nobody time to think. Is it worth $95 billion today? Probably not. Will it matter enormously in five years? Almost certainly.

Welcome to AI investing, where both things are simultaneously true.

GOOGLE PREPARES TO DROP GEMINI 4 ON THE WORLD AT I/O ON MAY 19 — AND APPLE BETTER BE WATCHING

📡 SOURCE: CNBC / Android Authority / Engadget

With Google I/O opening May 19, Google is set to unveil Gemini 4 alongside ‘Gemini Intelligence,’ a sweeping agentic AI system for Android that can operate across apps, understand the screen, and automate multi-step tasks — launching first on Samsung Galaxy and Pixel devices this summer, then expanding to cars, glasses, and laptops.

Keywords: Google I/O 2026, Gemini 4, Android AI, Gemini Intelligence, Google AI announcements

If there’s one company that has spent the last twelve months running like it borrowed someone else’s legs, it’s Google. And on May 19, when Google I/O 2026 opens its digital curtain, the search giant is expected to show the world just how fast it’s been sprinting.

The main event: Gemini 4. Google’s new flagship AI model is reportedly faster, smarter, and more deeply integrated across Google’s entire product universe than anything they’ve shipped before. In Gemini 4’s case, the specifics are genuinely interesting — a model designed from the ground up to reason across text, images, video, and audio simultaneously, without the awkward transcription middleware that made earlier multimodal systems feel clunky.

But the real story might not be the model itself.

It’s ‘Gemini Intelligence’ — Google’s agentic AI layer for Android that can move between apps, read your screen like a human, and execute complex multi-step tasks. Imagine telling your phone ‘book me the cheapest flight to Miami next Thursday and add it to my calendar’ and having it actually do it, without you touching a single button.

That’s the pitch. It’s enormous if it works.

The strategic urgency here is palpable and it has a name: Apple. Cupertino is widely expected to drop its own significant AI reboot soon, and Google is determined to make Gemini so deeply embedded in Android — in 250 million cars via Android Auto, on Pixel and Samsung devices, in Chrome, in watches, in glasses — that by the time Apple makes its move, Gemini feels like it was always there.

Whether Google can actually execute on this vision — instead of the usual ‘announced at I/O, quietly cancelled by November’ fate of many a Google product — remains the most important open question in consumer technology right now.

We’ll know a lot more after May 19.

ANTHROPIC NOW RICHER THAN FORD, GM, AND BOEING COMBINED — ALL THANKS TO GOOGLE’S STAGGERING $40 BILLION BET

📡 SOURCE: TechCrunch / Anthropic / CNBC

Anthropic’s run-rate revenue has surpassed $30 billion — tripling from $9 billion in just six months — fueled by Google’s $40 billion investment commitment in cash and compute, plus a new infrastructure partnership with Google and Broadcom designed to scale capacity for exponential demand growth.

Keywords: Anthropic valuation, Google Anthropic investment, Claude AI revenue, AI funding 2026, Anthropic Broadcom partnership

Somewhere in San Francisco, the ghosts of everyone who called Anthropic ‘the responsible AI company that was too cautious to compete’ are eating their words with a very large spoon.

The numbers coming out of Anthropic right now are genuinely staggering.

Run-rate revenue over $30 billion. That’s not a projection or a VC pitch deck with hockey-stick fantasies. That’s what their customers are actually paying, annualized. At the end of 2025, that number was $9 billion. They more than tripled in less than six months.

Companies spend entire decades trying to grow like that.

The fuel for this rocket ship? Google, which committed up to $40 billion in cash and compute — one of the largest corporate investments in any single AI company in history.

Combined with Anthropic’s new infrastructure partnership with Google AND Broadcom, this isn’t just a funding round. This is the construction of an entirely new AI industrial complex.

What’s particularly fascinating about Anthropic’s trajectory is how they got here. They didn’t compromise their safety-first positioning — they weaponized it.

In a market increasingly spooked by AI risks and hallucinations, Claude’s reputation for being the ‘thoughtful’ AI became a genuine enterprise selling point. Fortune 500 companies, law firms, healthcare systems flocked to the AI that wouldn’t embarrass them in a board presentation.

The irony is that Anthropic is now so large that questions about whether ‘safety-focused’ remains possible at $30 billion scale get genuinely interesting. But that’s a philosophical debate for another day.

Today, Anthropic is one of the most valuable private companies on the planet, and the scoreboard doesn’t lie.

STANFORD’S BRUTAL AI REPORT: MACHINES NOW BEAT HUMAN EXPERTS ON SCIENCE — BUT AI FIRMS ARE HIDING THE EVIDENCE

📡 SOURCE: Stanford HAI / IEEE Spectrum

Stanford’s 2026 AI Index report reveals AI models now surpass human expert baselines on PhD-level science benchmarks, generative AI hit 53% global adoption faster than any technology in history — and, troublingly, AI model transparency scores have collapsed from 58 to just 40 points as the most capable models disclose the least about themselves.

Keywords: Stanford AI Index 2026, AI surpasses human experts, AI transparency decline, generative AI adoption, AI benchmark performance

Stanford’s 2026 AI Index landed this week like a very dense textbook dropped on a very quiet library floor. It confirms everything AI optimists have been celebrating and everything AI pessimists have been losing sleep over — sometimes in the exact same paragraph.

Start with the headline finding: AI models now score above the human expert baseline on PhD-level science benchmarks.

Not ‘approaching.’ Not ‘nearly.’ Above. Google’s Gemini Deep Think won a gold medal at the International Mathematical Olympiad. AI systems are now solving at near-100% on SWE-bench Verified — a software engineering test that was at 60% just one year ago.

For years, experts debated when AI would match human experts. Apparently, that already happened while we were busy arguing about ChatGPT writing bad poetry.

The adoption numbers are equally mind-bending. Generative AI reached 53% population adoption in just three years. The personal computer took over a decade to get there. The internet took longer. Generative AI did it before half the people using it could clearly explain what a large language model is.

This is the fastest technology adoption curve ever recorded in human history.

But here’s where Stanford’s report gets genuinely alarming: as AI models get more powerful, they are becoming LESS transparent, not more. The Foundation Model Transparency Index dropped from 58 points to 40. The worst offenders are the most capable models. The companies building the most powerful AI are telling us the least about how it works, what data it trained on, and what its failure modes are.

At a moment when governments are trying to regulate AI, this is a five-alarm transparency crisis dressed in a business suit.

The workforce data is sobering too: entry-level software developer jobs for workers aged 22-25 have dropped nearly 20% since 2024. AI isn’t replacing all jobs — but it IS already eating the entry rung of the ladder for knowledge workers. The question of where young workers build experience that used to come from those entry-level roles remains loudly, dangerously unanswered.

80,000 TECH WORKERS FIRED IN THREE MONTHS — BOSSES BLAME THE AI, EXPERTS BLAME THE BOSSES

📡 SOURCE: Tom’s Hardware / CBS News / Washington Post

The tech industry cut nearly 80,000 jobs in Q1 2026 — the highest quarterly figure in at least two years — with nearly 50% attributed to AI-driven automation. AI became the top cited reason for layoffs in April for the second consecutive month, as Amazon, Snap, Coinbase, and PayPal announced major workforce reductions explicitly tied to AI adoption.

Keywords: tech layoffs 2026, AI job cuts, automation workforce, Amazon layoffs, PayPal AI restructuring

The quarterly layoff numbers are in and they are not great. Eighty thousand technology workers shown the door in three months, and nearly half of those companies have the audacity — or the honesty, depending on your mood — to blame artificial intelligence.

We have officially entered the era where ‘the algorithm did it’ is a termination letter.

The roster of companies citing AI while waving goodbye to workers reads like a who’s who of Silicon Valley: Amazon cut 16,000 corporate employees and pointed to ‘agentic AI workflows.’ Snap fired 1,000 people and specifically said AI lets smaller teams do the same work. Coinbase axed 14% of its workforce. PayPal is planning to cut 20% of its staff over the next few years — nearly 5,000 people — in a restructuring explicitly tied to AI automation.

Here’s where it gets philosophically messy: experts are split on whether AI is actually doing the cutting, or whether companies are using AI as a convenient scapegoat for layoffs that were coming anyway.

Harvard Business Review ran a piece headlined ‘Companies Are Laying Off Workers Because of AI’s Potential — Not Its Performance.’ Translation: bosses are firing people in anticipation of what AI might do, not what it’s actually done yet.

That’s a staggering indictment — essentially destroying livelihoods based on a technology forecast.

The real-world numbers tell part of the story: Goldman Sachs projects generative AI could ultimately affect 300 million jobs globally. AI was cited as the top reason for layoffs two months running. And the AI economic value is real — generative AI tools are estimated to be worth $172 billion annually to U.S. consumers.

Something is definitely happening. The debate is over who’s holding the knife.

What nobody is talking about enough: the 80,000 people who lost jobs this quarter aren’t abstractions in a white paper about productivity gains. They are engineers, coders, analysts, parents. The AI revolution is arriving with more receipts than apologies, and the social contract around work is fraying faster than policy can respond.

OPENAI HANDS BRUSSELS A ‘CYBER AI’ MODEL AS PEACE OFFERING — ANTHROPIC TELLS THE EU TO TAKE A SEAT

📡 SOURCE: CNBC

OpenAI announced it will grant EU regulators access to GPT-5.5-Cyber as part of cooperative oversight efforts — while Anthropic has refused to provide EU access to its powerful Mythos model, creating a rare public split between the two leading AI safety companies on the fundamental question of government oversight.

Keywords: OpenAI EU regulation, Anthropic Mythos model, AI government access, GPT-5.5-Cyber, EU AI oversight

In the ongoing saga of AI Companies vs. The European Union, this week served up a genuinely unexpected plot twist: OpenAI blinked first and actually handed Brussels a powerful AI model, while Anthropic — the company famously committed to safe and responsible AI — told EU regulators to sit and spin.

Let’s let that irony breathe for a moment. OpenAI, the company that has faced congressional hearings, FTC investigations, and a lawsuit from its own co-founder, is now voluntarily handing European regulators access to GPT-5.5-Cyber, a specialized cybersecurity variant of its latest model.

Meanwhile Anthropic, which built its entire brand identity around the idea that AI should be developed safely and transparently for humanity’s benefit, is keeping its Mythos model firmly locked away from EU review.

Somewhere in a philosophy classroom, a professor is updating their example of irony.

The strategic logic isn’t hard to decode. OpenAI’s regulatory cooperation play is clearly about building goodwill with a European market worth billions and greasing the skids for commercial expansion in a region historically suspicious of American tech giants. It’s regulatory judo: give regulators something to poke at before they force you to give them everything.

Anthropic’s refusal is harder to read. The company hasn’t offered a detailed public explanation. Possibilities range from genuine concerns about sharing proprietary model weights with government bodies, to competitive strategy — Mythos may be their most powerful model and they don’t want anyone, even regulators, poking around in it — to pure stubbornness. The EU’s AI Act gives regulators fairly broad powers to demand access to frontier models, so ‘holding out’ may not be a sustainable position.

The deeper issue is what this split reveals about the AI safety community’s complicated relationship with government oversight. Companies that built their reputations on transparency are now drawing lines around what governments get to see. That tension isn’t going away — and as AI models get more powerful, the line between regulatory cooperation and regulatory capture is going to get increasingly, dangerously blurry.