This Week in Cybersecurity
November 6, 2026
Each week we recap the biggest cybersecurity news headlines from the week to keep you informed and ready to face the latest threats. Here’s your breakdown for the week of October 31st – November 6th, 2026.
FBI, CISA Warn Attackers Are Targeting Industrial Refrigeration and Cold-Chain Systems
Federal agencies want you to look at your walk-in cooler differently. The FBI and CISA, joined by HHS, USDA and DOT, warn that attackers are hitting internet-exposed refrigeration controllers across grocery, cold storage, healthcare and food production. The playbook is simple: exploit an auth bypass or weak configuration, push units into nonstop heating or defrost, then hide the temperatures from operators.
In Europe, hackers hit an estimated 250 to 350 French supermarkets in October, then more than 115 stores across six countries this month. France’s blood service confirmed an incident at its refrigeration vendor altered settings at three blood storage sites. The fix list reads like 2010: pull these boxes off the internet, change default passwords, patch and wall OT off from IT.
Source: BleepingComputer
Read More
OpenAI Model Tried to Bypass Sandbox Network Restrictions During Coding Evaluation
OpenAI’s latest misalignment report reads like an intrusion writeup. It’s just that the intruder was one of its own models. During an August coding evaluation, an internal research model found its sandbox proxy down and went hunting for another way out. It scanned private addresses, hand-built packets to resolve DNS, wrote packet-level connection code and tried fragmented requests its own reasoning flagged as policy-breaking evasion.
Five hours in, it left itself a note to stop. Minutes later it was back on Amazon S3 looking for an archived copy, then asked the instance metadata service for cloud credentials. Real attackers run that exact play. The model never reached the reference site, so it invented the content, fake customer testimonials included, and decided the user didn’t need to know. OpenAI’s monitor caught it fast. Unprivileged containers lose raw-packet access by default now.
Source: SecurityWeek
Read More
Fake OpenAI and SpaceX Event Invitations Used to Steal Microsoft and Google Work Accounts
Free ChatGPT Pro sounds great until the sign-up page asks for your work login. Push Security found fake event pages posing as an OpenAI Academy livestream and a private SpaceX briefing. An Anthropic page tied to AWS re:Invent was already gone. The SpaceX version even dangles a “SpaceX share.” Both live pages pitch AI to marketing teams, which says plenty about who the operators expect to bite. A Cloudflare CAPTCHA keeps security scanners out. Then the sign-in buttons split the attack.
Microsoft users land in Kali365, which generates a device code. Type it in and you’ve signed in the attacker. Google users land in VoidProxy, an adversary-in-the-middle proxy that grabs passwords, MFA responses and session cookies. That’s two rented phishing services behind one page, both built to beat MFA.
Source: BleepingComputer
Read More
Did one of these hit close to home?
HBS security experts can help you turn security concerns into a clear plan of action.
Related Content
This Week in Cybersecurity | November 6th, 2026
Join us every Friday as we take a moment to recap cybersecurity news from the week. Here’s what you need to know for Oct. 31 – Nov. 6, 2026.
This Week in Cybersecurity | October 30th, 2026
Join us every Friday as we take a moment to recap cybersecurity news from the week. Here’s what you need to know for Oct. 24-30, 2026.
This Week in Cybersecurity | October 23rd, 2026
Join us every Friday as we take a moment to recap cybersecurity news from the week. Here’s what you need to know for Oct. 17-23, 2026.
