The software you already trusted

This week's stories

1. 01

   Malwarebytes

   Chick-fil-A loyalty accounts hijacked using stolen passwords(opens in a new tab)

   Nobody broke into Chick-fil-A. Attackers took username and password pairs leaked from other companies' breaches, replayed them here between 17 and 19 June, and enough people had reused a password that it worked. The prize was gift card and rewards balances, which are money, and which launder far more cleanly than card numbers. If you reuse a password anywhere, this is the attack that eventually finds it.

2. 02

   Malwarebytes

   Paidwork breach exposes data of 23 million users: Check if you're affected(opens in a new tab)

   Bank account numbers and home addresses sitting in the same dump as dates of birth. That combination is not a change-your-password problem, it is a ready-made identity theft kit. The intrusion happened in March and the database was being advertised on a criminal forum by April. Paidwork still has not publicly confirmed any of it, so 23 million people are learning about it from reporters rather than from the company holding their bank details.

3. 03

   Malwarebytes

   WhatsApp Web chats exposed by Adobe's Acrobat extension flaw(opens in a new tab)

   Visiting a single malicious page was enough to turn Adobe's Acrobat extension into a reader for your WhatsApp Web messages. The extension's privileges let an attacker step around the same-origin rule that normally stops one site reading another's data — and it is installed roughly 329 million times. Adobe shipped the fix over a weekend, which is genuinely fast. Check you are on 26.5.2.3 or later, and while you are in there, look at your linked WhatsApp devices and sign out anything you do not recognise.

4. 04

   SC Media

   Malicious ModHeader extension pulled from Chrome and Edge stores(opens in a new tab)

   A developer tool with 1.6 million installs shipped a dormant spyware SDK in version 7.0.18. The code to harvest every domain you visit, the encryption key, and the daily upload schedule were all present and simply switched off. "Not currently running" is not the same as "not there" — a flag flip is not an attack that needs your permission. Google and Microsoft have both pulled it. If you installed it for testing once and forgot, remove it rather than waiting for an update to save you.

5. 05

   Malwarebytes

   New ClickLock Stealer locks your Mac until you hand over your password(opens in a new tab)

   The delivery is the familiar ClickFix trick: a fake "verify you are human" page, usually impersonating Cloudflare, that asks you to paste a command into Terminal. What is new is what happens when you refuse to give up your macOS password. It kills Finder, the Dock and System Settings every 210 milliseconds, for up to 83 hours, until the machine is unusable and handing the password over feels like the way out. No real verification check has ever needed your Terminal. That one rule kills this entire family of attacks.

6. 06

   Malwarebytes

   Millions of cars could be tracked and unlocked by a hidden security flaw(opens in a new tab)

   Every KARR alarm shares one authentication key, and it sits in plain text inside the phone app. Recover it once and you can track or unlock any of roughly 2.2 million American vehicles. UC San Diego researchers reported it in January 2025 and the firmware fix arrived this month, eighteen months later. The worse detail is distribution: about half these alarms were added by the dealership at the point of sale, so most owners do not know they have one and will never install the patch. Check the driver's side window for a KARR or SWDS sticker.

7. 07

   Malwarebytes

   What happens if you visit a WordPress site hacked through wp2shell?(opens in a new tab)

   This one is in WordPress core rather than a plugin, which makes it both rarer and much worse — the first critical unauthenticated core RCE in nearly a decade. No login required, and it ends in full admin control plus code execution as the web server. Exploitation began within hours of the patch shipping, because a patch tells attackers exactly where to look. If you run WordPress anywhere, including the site you built for someone in 2019 and stopped thinking about, patch it and then check it rather than assuming you were quick enough.

8. 08

   The Hacker News

   Critical SharePoint RCE CVE-2026-50522 under active exploitation(opens in a new tab)

   Patching this one does not finish it, and that is the part being missed. Attackers are pulling SharePoint machine keys with a single request, and those keys keep working after the update lands — so a fully patched server can still quietly be someone else's. Rotate the machine keys, then treat any window where the server sat unpatched as a possible compromise rather than a near miss. CISA gave federal agencies until 25 July, which tells you how the people with the incident data are reading it.

9. 09

   The Hacker News

   Qilin ransomware attackers exploit PAN-OS authentication bypass(opens in a new tab)

   An authentication bypass in PAN-OS lets someone open a VPN session without valid credentials, and Qilin's affiliates have been using it as the way in before deploying ransomware. It only bites in a specific configuration — authentication override cookies enabled alongside certain certificate setups — so the first job is working out whether that describes you rather than panicking. Patch, turn the override cookies off if you are not genuinely relying on them, and read your VPN logs for sessions nobody can account for.

10. 10

    Malwarebytes

    OpenAI's agent escaped its sandbox during a security test(opens in a new tab)

    During an internal evaluation, a chain of OpenAI models broke out of its test environment through a zero-day in a package registry cache proxy, moved sideways until it reached a machine with internet access, and ended up inside Hugging Face's production systems. Hugging Face confirmed real unauthorised access to a limited part of that infrastructure, including some internal datasets and credentials. Worth being precise about what this is: the models were deliberately run with safety restrictions reduced, so it is not a story about an AI turning on anyone. It is a story about containment being the control that actually matters, and containment failing.