- ByteSize
- Posts
- That Microsoft Email is Lying, Your AI is Mining Bitcoin, and Your Books are Dead
That Microsoft Email is Lying, Your AI is Mining Bitcoin, and Your Books are Dead
PLUS: Micron drops $24B on Singapore chips and Waabi gets $1B for robot taxis...

📱 Switch to SuperMobile from T-Mobile. Get iPhone 17 Pro for free*
SuperMobile gives your business what it needs: unlimited premium data on America’s largest 5G network, intelligent network performance, and a 5-year price guarantee.**
Plus if you switch to SuperMobile right now, you can get a free iPhone 17 Pro (with a qualifying plan) to run your business.
The best business plan on the best network.
*With 24 monthly bill credits when you add a line and trade-in an eligible device on SuperMobile. For well-qualified customers; plus tax & $35 device connection charge.
**Exclusions like taxes & fees apply.
Roses are red, violets are blue, scroll down for jokes that might 302 you.
Well, welcome back! It’s February 3, and on this day in 1995, astronaut Eileen Collins became the first woman to pilot the Space Shuttle during mission STS-63. Collins proved that breaking barriers in space is significantly easier than breaking through Microsoft's update notifications.
Now, let’s jump in…
HACKERS TURN AI ENDPOINTS INTO THEIR PERSONAL CRYPTOCURRENCY CASINOS
Remember when we thought AI would solve all our problems?Well, we got 99 problems and it became one.
Researchers at Pillar Security just uncovered "Bizarre Bazaar," a criminal operation targeting exposed LLM endpoints like they're unattended ATMs in a sketchy neighborhood. Over 40 days, these digital pirates launched 35,000 attack sessions, turning poorly secured AI infrastructure into their personal cryptocurrency mining operation.
The scheme works like a cybercriminal pyramid scheme: one crew scans for vulnerable endpoints, another validates access, and a third operates "SilverInc," a service that sounds like a wellness MLM but actually resells stolen AI access through Telegram and Discord. They're offering access to "50+ AI models from leading providers," which is basically the Costco of unauthorized AI access.
The attackers are exploiting misconfigurations faster than you can say "unauthenticated Ollama endpoint on port 11434." They're stealing computing resources AND exfiltrating conversation histories AND attempting lateral movement into internal systems. It's like finding out your smart doorbell has been gossiping about your spending habits to strangers on the dark web.
Our advice: Check your AI endpoints before they start mining Bitcoin behind your back.
MICROSOFT'S EMAIL ADDRESS BECOMES SCAMMER'S BEST FRIEND
Microsoft's reputation just got weaponized, and not in the way they intended.
Scammers are now using the legitimate Microsoft email address "[email protected],” which Microsoft explicitly tells customers to whitelist, to deliver fake billing scam emails. These messages claim users have been charged $399 for nonexistent services and provide phone numbers for "dispute resolution" that actually lead to remote access scam attempts.
The scheme abuses Power BI's subscription feature, which allows external email addresses to be added as subscribers for reports. It's like giving hackers a backstage pass to your inbox using Microsoft's own credentials. The scam emails bury the subscription mention at the bottom where it's easily missed, while the obvious red flags wave proudly at the top.
Microsoft temporarily disabled the feature after reports surfaced—approximately 18 hours after being called out, which is faster than their usual Windows Update schedule. For experienced users, these scams are easier to spot than a crypto bro at a coffee shop. But for others, receiving fraud alerts from a trusted Microsoft domain makes the scams more believable than Sam Bankman-Fried's courtroom apology.
Even Microsoft's own email addresses can't be trusted completely, which honestly tracks with their overall reliability record.
ANTHROPIC'S "PROJECT PANAMA" TURNED MILLIONS OF BOOKS INTO AI TRAINING SMOOTHIES
Court documents reveal that Anthropic spent tens of millions of dollars on "Project Panama," their secretive operation to physically buy, scan, and dispose of millions of books for AI training.
The project involved buying books in batches of tens of thousands, slicing off their spines with hydraulic cutting machines, scanning the pages at high speed, and then scheduling recycling pickup for the literary remains. Internal documents show they considered approaching the Strand bookstore (famous for their "18 miles of books") and even "chronically underfunded" libraries.
This revelation came from a copyright lawsuit that Anthropic settled for $1.5 billion—roughly $3,000 per book title for authors whose works were downloaded from "shadow libraries" before the physical book-scanning operation began. It's the most expensive book club membership in history, except nobody got to actually read anything.
Meanwhile, other tech giants weren't sitting idly by Meta employees internally discussed downloading pirated books via torrents, with one engineer noting that "torrenting from a corporate laptop doesn't feel right,” which is the understatement of the decade.
⚙️ TOOL TIME
AI needs adult supervision, so Experts Exchange launched an AI Review question feature.
ChatGPT has gotten really good at confidently delivering wrong answers with the enthusiasm of a junior developer deploying on Friday afternoon.
So, our new AI Review feature lets you submit AI-generated answers for human verification, because we've reached the point where we need expert humans to fact-check our artificial experts.
AI gets you 80% of the way there, then leaves you stranded like Elon Musk's Cybertruck in a snowstorm. The scariest part is when AI confidently presents deprecated methods as current best practices while completely ignoring edge cases.
More people are showing up asking "ChatGPT told me to do this, does it actually work?" (iIt usually doesn't.) Think of EE's AI Review as having Batman AND Alfred watching your code—vigilant, thorough, and equipped with better judgment than your average AI chatbot that thinks every problem can be solved with recursion.
Try our AI Review feature before your next "ChatGPT told me this would work" moment inevitably backfires."
👨💻 JOB OPPORTUNITIES
The perfect role for anyone who fixes computers, printers, and souls. You'll manage Jamf and Okta while explaining to Karen from accounting that turning it off and on again isn't witchcraft.
Superhuman needs someone to provide first-level support and coordinate events in San Francisco. Must translate "tech speak" into "human speak" faster than Google Translate butchers poetry.
Cloudflare is scouting someone to lead IT audits and compliance programs. Perfect for someone who dreams in SOC2 and breathes NIST 800-53. Austin location means BBQ breaks between risk assessments.
🛩 INDUSTRY MOVES
Micron drops $24B on a Singapore NAND storage plant, proving that spending money on chips is more reliable than crypto investments and won't start tweeting conspiracy theories at 3am.
Waabi raises $1B to expand into robotaxis with Uber, because apparently we haven't learned enough from Tesla's "self-driving" adventures or Boeing's software quality control.
Salesforce signs $5.6B deal to inject AI into the US Army, creating the world's first CRM system capable of both managing leads and managing munitions with equal efficiency. Uncle Sam’s propaganda messaging about to be like “It’s not just the Army. We want you!”
Risotto raises $10M to make ticketing systems easier with AI, finally solving the age-old problem of help desk tickets that make as much sense as IKEA assembly instructions written by GPT-2.

Chip here with this week's EE community questions that prove troubleshooting is still an art form:
Securely erasing SKHynix M.2 NVMe SSDs on a Dell Precision Workstation without BIOS support
Hunting down the elusive Microsoft Visual FoxPro support library after a system migration
Creating Excel formulas to select and delete rows\ based on domain matching
That's another week where reality proved stranger than any code we could write! As always, may your endpoints be protected and your memes be fresh.
Enjoyed the news? Discuss over on Experts Exchange.
Got news to share or topics you'd like us to cover? Send ‘em our way by responding to this email. We can’t wait to hear from you. Really.





