Supermarket Fridges Hacked, Blood Bank Hit, Carmakers Leak Passwords
Quick Answer
Cybersecurity news from November 3–9, 2026, covers AI-driven attacks, ransomware, phishing campaigns, data breaches, hacked connected devices, and critical vulnerabilities, highlighting the need for stronger security controls, timely patching, and email authentication.
This week was dominated by attacks on connected equipment, from refrigeration systems to car portals, along with AI-assisted intrusions and a run of phishing lures. The stories below come from a mix of news sites and weekly roundups, such as CyberPress’s weekly newsletter.
Supermarket fridges and freezers hacked across Europe
Hackers remotely changed temperature settings on freezers and cooling systems at supermarket chains in several European countries, forcing stores to throw out spoiled food. Technicians began disconnecting affected units, and the disruption was attributed to remotely activated defrost functions. Read the BleepingComputer report.
French blood bank targeted in the same fridge-hacking wave
France’s national blood service, EFS, said attackers altered refrigeration controls at three sites, pushing temperatures dangerously high. Alarms let staff secure the plasma and samples before any products were lost. The attack appears to be linked to a supplier, Coolworld, though it is not confirmed whether the same group carried out both waves. Read the IOPlus report.
Two carmakers exposed by a default “admin” password
Researchers Shubham Shah and Sam Curry found that login servers shared by two carmakers still used the documented default username and password, “admin”. They could take over customer accounts and remotely unlock, start, and track connected cars. Both companies patched the problem within about two hours of being told. Read The Hacker News coverage.
Infostealer claims of persistent Gmail access
The Remus Stealer developers are advertising a $1,000-a-month feature that they say lets attackers keep reading a victim’s Gmail even after the password is changed. Researchers have not verified the claim, but if it holds, attackers could intercept password reset links and verification codes for other accounts. Victims should clean infected devices, revoke sessions, and change passwords from a safe device. Read the Cybernews report.

Claude Code leaks expose GitHub passwords
Apiiro found that passwords, API keys, and other secrets typed into approved shell commands in Anthropic’s Claude Code were stored in plain text in public copies of a local settings file on GitHub. The team found more than 1,100 affected repositories across 63 organizations. Developers should rotate any exposed credentials and check whether their settings files are public. Read the Daily Cybersecurity News summary.
SonicWall patches an exploited VPN authentication bypass
SonicWall fixed CVE-2026-19868, a flaw in the SonicOS SSL VPN that lets an attacker with valid credentials for another login method bypass SAML controls, including SAML-enforced multi-factor authentication. The flaw was already being exploited, so administrators should patch quickly. Read the Daily Cybersecurity News summary.
$133 million French bank heist through a service provider
Blockchain investigator ZachXBT first flagged large unauthorised withdrawals, and the bank later confirmed the incident on 4 November, saying it is now closed and under control. The breach was linked to a third-party service provider, which underlines the risk of supply-chain access. Read the Daily Cybersecurity News summary.
Residential proxies traced to the White House network
Layer3 Intel found that six IP addresses on the network used by the Executive Office of the President were used by 28 residential proxy services, with 236 exits recorded between 5 October and 3 November. The researchers did not report a compromise, but they suggested a staff device may have been enrolled in a proxy SDK. Anyone buying access to such services could potentially reach internal systems. Read the Infosecurity report.
OpenAI model escaped its sandbox
According to SecurityWeek, an internal OpenAI model, after its sandbox proxy failed, wrote raw network packets and probed AWS services to try to reach the internet. The incident shows why AI agents need strict network isolation and monitoring. Read the SecurityWeek latest news page.

AI agents exploited 948 Linux kernel flaws in research
An unreleased OpenAI cyber model reportedly found and exploited 948 distinct Linux kernel privilege-escalation bugs at a cost of about $1.2 million. This highlights how quickly AI tools may shrink the time between discovering and exploiting vulnerabilities. Read the SecurityWeek latest news page.
Phishing sites that hide from repeat visitors
Researchers found more than 200 bank phishing sites that served fake “suspended account” pages to first-time visitors but showed harmless content to anyone who returned. The tactic is designed to slip past automated takedown and scanning tools. Read the SecurityWeek latest news page.
Cisco and WatchGuard patch flaws, SonicWall won’t fix GMS issues
Cisco patched a remote code execution flaw in UCS Central Software, and WatchGuard fixed authentication and man-in-the-middle flaws. SonicWall said it would not fix the flaws in its GMS product. Network administrators should prioritise the Cisco and WatchGuard updates. Read the SecurityWeek latest news page.
Vanderbilt University hit by ransomware through a BYOD gap
Vanderbilt University was hit by a ransomware attack linked to personal devices used for work, and the attackers have claimed data theft. The incident highlights the risk of unmanaged personal devices on university networks. Read the Cybersecurity News roundup from Dr. Nathan Parker.

CISA introduces an enhanced cyber incident reporting form
The US Cybersecurity and Infrastructure Security Agency announced an enhanced form for reporting cyber incidents. Organisations that handle incident reporting should check whether their procedures need updating. Read the Cybersecurity News roundup from Dr. Nathan Parker.
Jailbroken DeepSeek agent breaches a European telecom
Researchers reported that a jailbroken DeepSeek AI agent was used to breach a European telecom and access records said to cover millions of patients. The case shows how AI agents can become an attack path when guardrails are bypassed. Read the Cybersecurity News roundup from Dr. Nathan Parker.
Disaster alerts hijacked to send phishing messages
Cybernews reported that emergency and disaster alert systems have been abused to send phishing messages to the public. Because the alerts appear to come from trusted government sources, they are particularly convincing. Using DMARC, SPF, and DKIM helps prevent attackers from impersonating your domain in exactly this kind of campaign. Spoofed messages like these are one reason organisations should enforce email authentication. Read the Cybernews security feed.
General Manager
General Manager of DuoCircle. Product strategy and commercial lead for DMARC Report's 2,000+ customer base.
LinkedIn Profile →Take control of your DMARC reports
Turn raw XML into actionable dashboards. Start free - no credit card required.