TechSnark Weekly #6
This week's news to lose faith in technology, but especially in humankind
Let's get started!
News n°1
The company that was supposed to protect civil liberties is now helping to... do the exact opposite. Welcome to the Palantir paradox, where employees are starting to wonder if they’ve accidentally ended up on the wrong side of history.
The betrayal of the original mission
Palantir was born with CIA funding after 9/11, with a goal that sounded almost noble: defending civil liberties. Eighteen years later, with Trump back in the White House, the company has become the DHS’s technological infrastructure for deporting immigrants. A former employee summarized the 180-degree turn with elegance: the threat “is now coming from within,” and the company that was supposed to prevent abuses “now seems to enable them.” In short, they went from defenders to executioners, just with better paychecks.
When the company chat gets awkward
In January 2026, a nurse named Alex Pretti was killed during protests against ICE. Palantir employees started asking uncomfortable questions. Management’s response? Automatically deleting Slack conversations after seven days in the #palantir-in-the-news channel. The official reason cited “information leaks,” but translated, it means: it’s better not to leave a paper trail of collective unease.
Things got worse on February 28, when a missile strike on an Iranian elementary school killed over 120 children, and it turned out that surveillance systems like Palantir’s Maven were used. At that point, even Slack’s auto-cleanup wasn’t enough to contain the discontent.
The manifesto of cognitive dissonance
During an AMA session in February, a member of the privacy team candidly admitted that “a sufficiently malicious client is practically impossible to prevent right now” and that attempts to talk sense into CEO Alex Karp have been “largely unsuccessful.”
Karp, meanwhile, published a manifesto based on his book in which he proposes reintroducing the draft. Employees complained because such statements make it harder to sell the software outside the U.S. Karp’s response was a masterpiece of CEO rhetoric: “if you have a position that never costs you a single employee, it’s not a position.” In other words: if you don’t like it, there’s the door.
It’s hard to find a more “elegant” way to tell your employees that their ethical concerns are worth less than the corporate vision. And so, the company that was supposed to defend freedom continues undeterred to do the opposite, while those working there discover that maybe, after all, they really are the villains of the story.
Source: https://attackofthefanboy.com/tech/palantir-dropped-its-infamous-manifesto-a-while-back-and-now-its-employees-are-starting-to-question-whether-theyre-the-bad-guys/
News n°2
Meta and Microsoft have just given a crystal-clear lesson on how AI capitalism works in 2026: profits are soaring, Wall Street loves the stock, and those heading out the door are simply the “transition cost” toward a future where an Excel sheet carries more weight than any “talent.”
Meta is laying off 8,000 people (roughly 10% of its workforce) and canceling another 6,000 open positions, all effectively cut as of May 20.
Microsoft, on the other hand, avoids using the word “layoff”: it is offering the first “voluntary retirement” program in its history to up to 8,750 U.S. Employees, about 7% of its staff, provided that their age + years of service equal at least 70.
The result is identical: up to 23,000 roles are either vanishing or will never be filled, even though both companies are posting record-breaking revenue and profits. Economic hardship has nothing to do with it; quite simply, they’d rather spend their money elsewhere.
The real budget: machines, not people
Meta expects to burn between $115 and $135 billion in capex in 2026, nearly double the $72 billion spent in 2025, virtually all of it destined for data centers, Nvidia GPUs, custom chips, infrastructure for Llama, and the new Meta Superintelligence Labs division.
Meanwhile, in the fourth quarter of 2025 alone, it posted a net profit of $22.8 billion on $201 billion in annual revenue: no sign of a company “struggling.”
Microsoft follows the same logic, just a more polite version: quarterly revenue at $81.3 billion (up 17% year-over-year), Azure at +33%, with AI services accounting for 16 points of that growth, and hundreds of billions promised to data centers, Copilot, and the OpenAI partnership.
The generous exit packages for senior staff become a mere statistical detail in an AI spending plan that turns cut salaries into background noise.
Meta is proceeding with blunt layoffs across all business units while shifting engineers toward a new Applied AI division and toward its small-business advertising arm.
The corporate narrative has slid from “we have too many low performers” (2025) to “matters of efficiency and contribution,” which implicitly means: we’re not kicking you out because you’re bad at your job, but because you’re in the wrong department for the story we need to tell investors.
Microsoft, more refined, hides behind the “Rule of 70,” which effectively targets those in their 50s and 60s who built the pre-AI Microsoft.
The hardest-hit departments are precisely those where automation via Copilot and AI agents is most advanced (Cloud, Gaming, Global Sales), following a 2025 spent ramping up performance pressure and multiplying PIPs.
Outside of Meta and Microsoft, the pattern is the same: Oracle has cut up to 30,000 roles (about 18% of its workforce) to shift $8-10 billion a year toward a $156 billion AI infrastructure plan, while its remaining contractual obligations reach $523 billion and profits have nearly doubled.
Amazon has restructured 16,000 positions, Dell 11,000, and Snap has cut 16% of its staff: in 2026, over 96,000 tech workers have already lost their jobs, 40% more than in the same period in 2025.
Since 2020, tech layoffs have neared 900,000 positions, with a narrative that has evolved with suspicious convenience: first “we over-hired during Covid,” then “we are restructuring for AI,” and now “we are cutting for AI”, stated explicitly, as if it were an act of modernization.
It has become a sort of magic formula: say “AI” and everything, from staff cuts to collapsing internal morale, suddenly becomes rational and inevitable.
The trendy term is “AI-washing”: layoffs announced in the same breath as a solemn nod to AI strategy, without any proof that internal systems are actually capable of doing the work of the people being sent home.
New divisions like Superintelligence Labs and Meta Compute exist in press releases far more than in operational processes, but they are enough to create the illusion that Meta’s $115-135 billion in capex or Microsoft’s $81 billion in annual investments are “naturally” replacing thousands of salaries.
The bottom line: it’s not the technology, it’s the hierarchy of values
From 2022 to 2026, we’ve seen three waves: first the post-pandemic excuse, then restructuring “for AI,” and now the direct, unapologetic link between cuts and AI investment.
Oracle openly says it lays people off to build data centers, Meta says it cuts to “offset” AI investments, and Microsoft designs a program tailored for those who are no longer aligned with its AI-centric future.
The slaughter has only just begun.
Source: https://thenextweb.com/news/meta-microsoft-layoffs-23000-ai-spending
News n°3
Putin has found a new way to get under Russians’ skin: intermittently shutting down the internet, especially mobile data, under the guise of security and the war effort.
The result: people can’t work, communicate, or access basic services, and they are suddenly realizing that the promised “stability” comes with a very concrete price tag in their daily lives.
It’s no longer just dissidents and activists complaining; now it’s influencers, starlets, and typically apolitical figures who, until yesterday, were hawking face creams and a patriotic lifestyle, but today are openly saying that life in Russia has become unlivable.
One of their posts goes viral, forcing the Kremlin to respond and promise vague “course corrections,” as if the whole thing were just a minor technical glitch.
Meanwhile, the state is pushing citizens toward the national super-app MAX, controllable and unencrypted, which is a great idea if your goal is to turn everyone’s phone into an extension of the censorship office. VPNs, workarounds, hacks: many Russians are trying to fight back, but between blackouts and throttling, even digital resistance has its limits.
Official polls, the ones that usually show undying love for the leader, are starting to dip: Putin’s approval rating is sliding back toward pre-invasion levels, while irritation and fatigue over the war, taxes, and online blocks continue to grow. Messing with the internet, essentially hitting people where they live, seems to hurt his approval ratings more than a thousand patriotic TV appeals ever could.
Even the “controlled opposition” is waking up: Communists and new pro-government parties are openly criticizing the blackouts, feigning surprise at a system they themselves have legitimized for years.
Some venture comparisons to Tsarist Russia, speaking of growing discontent and the risk of “more decisive steps,” but in practice, no one is truly challenging the top tier of power.
Underneath it all remains a country where more and more people view elections as a charade, speak of Russia as an open-air prison, and no longer believe the Kremlin’s promises to fix things.
For now, the regime isn’t collapsing, but it has discovered an uncomfortable truth: you can control the media, but once you start cutting the connection, even the faithful begin to wonder if you’ve finally gone too far.
Source: https://www.nytimes.com/2026/04/28/world/europe/russia-internet-restrictions-putin.html
News n°4
This isn’t some epic struggle over the fate of humanity; it’s just a run-of-the-mill settling of scores between two ultra-wealthy men accusing each other of being greedy, cynical traitors to the cause of “the greater good.” Honestly, it’s better than anything on Netflix.
OpenAI was founded in 2015 as a “different kind of lab,” a non-profit project designed to keep Artificial Intelligence from falling entirely into the hands of predatory corporations, driven by the rhetoric of “we’re doing this for humanity, not for profit.” Elon Musk provides the seed money, Sam Altman plays the charismatic leader, and on paper, the idea is noble: a sort of Manhattan Project for AI, but without the part where anyone makes an obscene amount of money. Too bad that right from the jump, Altman writes to Musk suggesting they bump researcher salaries by $100,000 to $200,000 “to signal that we’ll take care of them.” In other words: sure, it’s a non-profit, but with luxury startup pay.
From there, the writing was on the wall. The “non-profit” status lasted about as long as a press release. OpenAI transitioned to a for-profit structure, controlled by employees and institutional investors, Microsoft, Amazon, Nvidia, SoftBank, and a linked foundation, and is now gearing up for one of the biggest IPOs in history, with a valuation nearing a trillion dollars. Altman keeps repeating that he has no direct equity in OpenAI, yet he remains comfortably a billionaire thanks to other investments. Meanwhile Musk, the man portraying himself as the betrayed altruist, is doing exactly what he claims to loathe: building his own AI venture, xAI, within a highly profit-oriented company like SpaceX, which is also destined to go public.
The lawsuit is, officially, a matter of principle. Musk accuses Altman and President Greg Brockman of “unjustly enriching themselves” by billions while betraying OpenAI’s original charter, which spoke of benefiting humanity rather than lining the pockets of a few individuals and major shareholders. OpenAI fires back like they’re in a high-end HOA dispute: you’re not the good guys, he’s the greedy one. The company claims Musk abandoned the project only after realizing he wouldn’t be able to seize total control. Both sides compare themselves, with admirable modesty, to a Shakespearean drama: Musk speaks of “perfidy and deceit of Shakespearean proportions,” while Altman notes that the AI sector is now full of “Elizabethan tragedy-level drama.” The explicit reference is to Julius Caesar: the friend who stabs you in the back, power that corrupts, except here, the blood is replaced by emails and stock options.
The backdrop for all this is the 2026 version of the Regulatory Wild West. Congress is essentially paralyzed, federal watchdogs are weakened, and the Trump administration is packed with people very receptive to the money and glamour of Silicon Valley, the President included. In the absence of effective laws, civil courts are all that’s left: lawsuits become the only way to squeeze out some information and put even a semblance of a boundary on tech power.
Tech regulation activists view the situation with mixed feelings. On one hand, many want OpenAI to be held accountable for its choices just as much as Meta or YouTube. On the other, there’s very little faith in turning billionaires’ personal vendettas into the primary engine for “legislation via private litigation.”
The irony is that, depending on who wins, the outcome is equally bleak. If Musk prevails, OpenAI could be severely weakened or even dismantled, opening up a massive market share for one of Musk’s other companies to “swallow up” without much trouble. If OpenAI manages to get the case dismissed, the takeaway would be just as educational: it’s perfectly fine to start as a reassuring non-profit, promise to work for the public good, and then cynically flip into a trillion-dollar for-profit corporation with zero serious consequences.
Outside the courtroom, the tech world keeps racing: OpenAI is scaling projects ahead of its massive IPO, Google is pledging tens of billions to Anthropic, chipmakers are seeing double-digit growth, and AI is being shoved into everything from retail stores to prison chatbots. Inside the courtroom, however, old 19th-century statutes and apocalyptic fears about the future of AI clash in a trial that won’t make anyone a better person, but might at least force us to look at the structural hypocrisy of an industry that pretends to save the world while filling out an Excel sheet of its profits.
Source: https://www.nytimes.com/2026/04/28/technology/elon-musk-sam-altman-trial.html
News n°5
Yet another miraculous “document reader” has popped up on Google Play, one that did very little reading of documents but could certainly read plenty of cash right out of your bank accounts. Hidden behind that harmless icon was Anatsa, one of the most persistent Android banking Trojans of recent years. The app, published under the package name com.groundstation.informationcontrol.filestation_browsefiles_readdocs, easily blew past the 10,000-installation mark before anyone at Google decided to wake up and pull the plug.
Anatsa is hardly a new kid on the block; it’s been lurking since 2020, specializing in stealing credentials, logging keystrokes, and orchestrating fraudulent transactions while you think you’re just checking your balance. In the meantime, it has expanded its “client portfolio,” managing to target over 831 financial institutions worldwide, including banks and crypto platforms in countries like Germany and South Korea. The latest round was unmasked by the Zscaler ThreatLabz team, who found the app on the Play Store, analyzed it, and published their report in late April 2026, by which time, for the users, the damage was already done.
The strategy is simple, and it works precisely because it leans on a convenient assumption: “if it’s on Google Play, it must be safe.” The app arrives on the store seemingly clean: no embedded malicious payload, no suspicious behavior during the review process. Once installed, it acts like a normal file/document reader just long enough to earn your trust, while behind the scenes, it contacts a remote server and downloads the real cargo: the Anatsa Trojan, installed silently, without any theatrical pop-ups. It’s all thanks to a two-phase delivery designed specifically to bypass static analysis performed only on the app’s initial version.
Once the payload kicks into gear, its first move is to ask for accessibility permissions, masquerading them as necessary for the app to function properly. With that authorization, the malware gains a disproportionate level of control: it can overlay content on other apps, read SMS messages, show full-screen windows, and generally insert itself into any “sensitive” interaction without being noticed. This is how they get to your banking credentials: the user opens their banking app, and Anatsa places a fake login form on top, visually identical to the original, collects the username and password, and ships them off to command-and-control servers.
To top it off, Anatsa also checks its surroundings: if it detects an emulation or sandbox environment, it sticks to showing a perfectly respectable file manager interface, never revealing its criminal side. The result is that automated analysis systems see a harmless app, while on real smartphones, the Trojan goes about its business undisturbed.
In hindsight, the advice is the same stuff that’s been repeated for years but continues to go ignored until something goes wrong. An app pitching itself as a document reader has zero credible reason to ask for accessibility or SMS access; if it does, the correct response is “no,” followed by an immediate uninstall. Keeping Google Play Protect active, avoiding unknown developers, and being wary of “magic” utilities with five stars and two lines of generic reviews is the bare minimum. Anyone who installed the offending app should remove it immediately and scan their device with a decent security tool, not the usual ad-riddled “cleaner” junkware.
Source: https://cybersecuritynews.com/fake-document-reader-on-google-play/
News n°6
Ah, browser extensions. Those incredibly convenient apps that promise to block ads, enhance Netflix, or automate your job search. Too bad that while you think you’re getting a free service, they’re making the deal of the century by selling every single detail of your digital life to anyone willing to pay. And yes, it’s all perfectly legal.
LayerX Security did the dirty work of analyzing the privacy policies of thousands of extensions, discovering that at least 82 of them, with over 6.5 million users, are happily turning your data into cold, hard cash. But here’s the kicker: that’s only what they could actually verify, because 71% of the extensions in the Chrome Web Store don’t even bother to publish a privacy policy. Translation: they only analyzed 29% of the total, which means the number of extensions selling data could be in the tens of thousands.
Take the QVI empire, a network of 24 extensions for streaming platforms like Netflix, Disney+, and HBO Max, published by HideApp LLC. With 800,000 unsuspecting users, they collect viewing history, content preferences, age, and gender, only to resell the full package to movie studios and marketing agencies. Essentially, while you’re enjoying your favorite show thinking you’ve found a useful tool, they’re monetizing your TV habits.
But the real irony lies with the ad blockers. Eight of them, with over 5.5 million users, sell you the illusion of blocking ads while selling your browsing data under the table. Stands AdBlocker, with its 3 million users, sells everything for “market analysis.” Poper Blocker goes one better: with 2 million users, it doesn’t just sell identifiers and browsing activity, but also inferred sensitive data like health conditions, religious beliefs, and sexual orientation. So, you’re paying with your most intimate data just to avoid seeing banner ads. Brilliant.
And it doesn’t stop there. There are another 50 extensions with over 100,000 total users playing the same game. Career.io Job Auto Apply, for example, promises to automate job applications but then sells your resume data to data brokers. The tool that’s supposed to help you find a job is actually going on a shopping spree with your professional history.
The trick is always the same: buried inside those privacy policies that nobody ever reads, because let’s face it, who has time to sift through 47 pages of legalese? There’s always a little phrase like “we may sell or share your personal information with third parties.” Click “accept,” and voilà: everything becomes legal. You don’t even need to be particularly sneaky; you just have to be boringly bureaucratic.
The problem gets worse when these extensions end up on work computers. Twenty-nine of the 82 identified extensions are B2B sales intelligence tools that sell data. Installed on work machines, they can funnel internal URLs, SaaS dashboards, and corporate searches into commercial databases that your competitors can simply buy. Imagine the scene: your sales team installs an extension to find leads faster, and six months later, the competition knows exactly which clients you’re wooing because they bought that data from a broker.
LayerX advises organizations to implement extension governance policies, centralized management, and to consider blocking those that sell data or lack privacy policies. Sound advice, sure. But the real tip should be simpler: if a service is free, you’re probably the product.
Source: https://layerxsecurity.com/blog/your-extensions-sell-your-data-and-its-perfectly-legal/