Welcome to the weekly pileup of things that could ruin your day, your weekend, or your weekend’s weekend. The headlines are many, but the moral is boring and beautiful: attackers love ordinary things that are trusted too much. Extensions, APIs, packages, and services with too much freedom are the recurring villains.

Weekly headlines (fast, loud, slightly sarcastic)

  • Browser extensions impersonating crypto tools harvested session tokens and wallet data — sneaky add-ons pretending to be helpful turned out to be pickpockets.

  • AI-powered intrusion toolchains were used to coordinate attacks across several countries, turning reconnaissance and exploitation into an assembly line.

  • The U.K. warned about the dangers of employees using unapproved AI services — aka “shadow AI” — which can leak secrets and widen your attack surface overnight.

  • Scammers posing as execs pushed M&A conversations off official channels to WhatsApp and personal email, then tried to trick legal teams into wiring money.

  • Microsoft added age-awareness APIs to Windows 11 so apps can know whether they’re talking to a child, teen, or adult without begging for birthdays.

  • A gargantuan fake-shop network used over 119,000 domains to clone real stores and steal payment card details — the internet’s worst garage sale.

  • Browser makers tightened release cycles to ship security fixes faster because LLMs help find bugs faster — good bugs, bad timing.

  • Google pushed patches for about 200 Android vulnerabilities this month, including critical Wi‑Fi memory flaws. If you haven’t patched, assume you’re on the menu.

  • Two arrests in Singapore after a scheme abused mobile-shop access to hijack national authentication accounts and spin up payment accounts.

  • An email-provider breach exposed newsletter lists and spawned phishing that tried to trick hardware wallet users into handing over backups.

  • Tens of thousands of Plex servers are still exposed and vulnerable after recent disclosures — please stop hosting media servers like they’re Fort Knox.

  • A supply chain/malware chain delivered a rat, an AI-made backdoor framework, and then ransomware — a malware combo platter we all fear.

  • CISA refreshed its insider-threat guidance to reflect remote work, AI manipulation, and modern access risks.

  • A U.S. court handed down a long sentence for producing and weaponizing non-consensual AI porn — a grim reminder people use tech without ethics.

  • An alleged bank-takeover developer was extradited in connection with a large account‑takeover ring that used spoofed domains and fake login pages.

  • A 245-million-dollar crypto theft ringleader pleaded guilty — social engineering plus hands-on operations still pay, literally.

  • Teams will obscure external QR images by default to stop drive-by QR scams — tiny UX change, decent win.

  • Phishers started chaining legitimate cloud services to build trust paths that land victims on credential harvesters or drop remote access tools.

  • Smart TV data collection claims prompted scrutiny after a report said TVs were scanning networks and enumerating nearby devices; manufacturers pushed back.

  • Malicious npm packages masquerading as wallet libraries were found — if your dev machine installed one, assume compromise and rotate keys from a clean box.

  • Attackers began building phishing pages directly inside victims’ browsers using blob URLs, so there’s no persistent URL to block.

  • Over 5,400 small-business websites have been compromised to serve crypto-themed attack payloads that try to trick visitors into running commands.

  • Executive Social Security numbers keep appearing on underground markets — leadership data is prime shopping-cart material for fraudsters.

  • Large studies of AI tool collections found many tools can access sensitive data or execute commands; natural-language instructions themselves can become an attack vector.

  • Phishing-as-a-Service tooling that bypasses MFA hit organizations in 40+ countries, harvesting session cookies and full authentications.

  • The FBI published a new cyber strategy outlining disruption, attribution, partnerships, and the planned use of agentic AI for defensive operations.

Takeaways you can actually use (no buzzwords, just fixes)

Same lesson, repeated until it sticks: stop trusting everything by default. The problems weren’t magic — they were permissions, chains, and stuff nobody checked twice.

  • Least privilege — Trim access for extensions, services, and tools. If an add-on asks to read all your crypto pages, treat it like a stranger asking to sleep on your couch.

  • Patch. I know it’s a boring ritual, but 200 patched Android flaws aren’t helpful if you ignore updates.

  • Vet third parties and rotate secrets after any suspicious supply-chain or mailing-list incident. If your email provider is breached, assume the bad guys got a guest list.

  • Monitor exposed services (hello, Plex). If something is internet-facing, assume it’s probed within minutes — protect, isolate, or pull it offline.

  • Harden dev pipelines: audit packages before adding them, run builds in isolated environments, and rotate keys from uncompromised machines.

  • When in doubt, ask for proof on official channels. If someone pushes you to WhatsApp or a personal address for big-money transfers, stop and confirm.

  • Treat AI tools like colleagues that need onboarding: control what tools can do, inspect their prompts and outputs, and don’t let them inherit secrets by default.

Bottom line: attackers don’t need a castle siege. One unlocked side door, one overly-trusting service, or one lazy hinge is enough. Remember the boring boring stuff: limit what you give away, check who’s allowed to act, and patch the things that bite.