Daily OT Security News: November 16, 2026
Today’s briefing covers a German hotel pool where the chlorine alarm went out by email and nobody saw it during the Microsoft 365 outage, active probing of end-of-life Copeland refrigeration controllers within two hours of disclosure, a suspected malicious software update at drone maker Quantum Systems, October ransomware figures that put industrial manufacturing third on the victim list, and Marlink’s argument that offshore energy connectivity has to be engineered for resilience. Publication dates and event dates are separated below, and details that rest on reporting rather than an official statement are identified as such.
1. A hotel pool alarm went to email, and the email never arrived
Source and event dates: The incident occurred on Friday, November 13, 2026, at Hotel Ling Bao at the Phantasialand park in Brühl, Germany. WDR published the park’s account on November 14, and Der Spiegel, Welt and the Kölner Stadt-Anzeiger reported the same day. Golem followed on November 15 and Security Affairs on November 16. Microsoft described the cause of its outage in a statement on November 13.
News briefing
Fifteen guests at the hotel suffered respiratory irritation after using the pool, according to WDR. Thirteen complained first and the total rose to 15. Four guests were taken to hospital, and one stayed for observation with a moderate injury. Emergency services sent around 40 responders. The pool was closed, the water was replaced, and the fire service confirmed to WDR on Saturday that it had reopened.
Tests by emergency crews found a defect in the pool’s automatic chlorine dosing system. The system had put a substantially excessive amount of chlorine into the water. That defect, not the email failure, produced the chemical exposure.
The warning is where the cloud dependency matters. The park told WDR that its maintenance safety system for the sauna area had already raised an alarm by email. No email arrived during the Microsoft 365 disruption, so nobody saw the alert. In other words, the monitoring system worked and the notification path did not. WDR’s reporting is explicit that the emails did not arrive during the outage and that this is how the alarm went unnoticed.
Microsoft said the November 12 outage was a technical issue tied to an earlier network capacity upgrade, not a cyber or AI event, and marked the Exchange Online incident, EX1550543, resolved on November 13. Its threat intelligence team also reported a phishing campaign that posed as Microsoft 365 service health notices and referenced the real incident identifier. Some messages told recipients that mail held during the outage would be deleted within 24 hours unless they clicked a link. Microsoft named three sender domains and said the activity was consistent with credential phishing, with collection and malware delivery unconfirmed.
OT implications
A safety notification that travels through a third-party email service inherits that service’s availability. The alarm chain here had a single path: sensor, controller, mail relay, cloud mailbox, human. When any link failed, the alarm existed only in a log. Designs for occupied buildings and plants usually need a local annunciation path that does not depend on the internet, such as a panel alarm, a horn or strobe, or an on-premises gateway with its own connectivity. A watchdog that reports a failed send is also worth more than a green dashboard icon.
The dosing system also failed unsafe. It continued to add chlorine, and the only barrier was a notification that nobody received. Where a control loop can create a hazard, the safe state should be enforced in the controller or in independent hardware, not in the monitoring layer. Dose limits, runtime limits and a physical interlock are the controls that stop the event; the alarm only shortens the exposure.
Most sites test whether a sensor triggers an alarm. Fewer test whether the message is delivered, to whom, and how quickly. That end-to-end test is the one that would have caught this. The phishing angle compounds the problem, because outages create a credible pretext for messages that ask people to click. Service status should be checked in the admin center, not through a link in an email that cites an incident number.
Sources: WDR: IT-Panne Schuld für Chlorgasunfall im Phantasialand; Der Spiegel: 15 Verletzte wegen Microsoft-Ausfall; Security Affairs: The Microsoft outage left 15 hotel guests poisoned by chlorine; SecurityWeek: Microsoft rules out cyberattack in outage, warns of phishing campaign.
2. Copeland refrigeration controllers drew probes within two hours of disclosure
Source and event dates: VulnCheck’s CVE records for the 14 flaws were published on November 12, 2026. Exploitation attempts were recorded the same evening and on November 13. SecurityWeek reported the probing on November 16. The October 11-12 supermarket refrigeration attack and France’s November 9 attribution are background, not part of this disclosure.
News briefing
VulnCheck disclosed 14 vulnerabilities in the XWEB EVO refrigeration supervisor, a Dixell product now sold under the Copeland name. Researcher Valentin Lobstein demonstrated each flaw against appliance labs rebuilt from the vendor’s firmware, and each one produced root privileges. Nine of the flaws affect only firmware builds that predate a 2021 rewrite. They include unauthenticated command execution through a setuid-root CGI, command injection in the login page before any password check, a hardcoded Administrator password, and an unauthenticated write to the system configuration file.
The other five reach the current 5.8.8 firmware. Four involve predictable credentials. The Administrator’s daily password is generated from the device’s MAC address and the current date with no server-side secret, and an unauthenticated endpoint discloses that MAC address. The root password is built from the same daily code, so an attacker who can read the MAC can calculate the credentials offline, enable SSH and log in as root. The fifth is a command injection that requires authentication.
There will be no patch. Copeland ended software support for the XWEB EVO family on March 16, 2026 and has told customers to move to XWEB Pro. SecurityWeek reports that Onyphe data cited by VulnCheck shows nearly 2,000 internet-exposed XWEB EVO hosts, against 108 for XWEB Pro.
The reaction was fast. CVE records went live on November 12, and VulnCheck’s sensors saw attempts against CVE-2026-124014, the unauthenticated configuration write, within about two hours. One IP address sent 125 requests overnight, and on November 13 a second address sent four probes aimed at CVE-2026-124010, the unauthenticated root command execution. Both targeted flaws affect only the older firmware. Lobstein said neither scanning source is evidence of the October campaign that France attributed to Iran’s IRGC unit Shahid Kaveh, describing the October operation as targeted sabotage and the new activity as opportunistic hosts acting on fresh advisories.
OT implications
End-of-life equipment with no patch path is a permanent exposure, and the only remaining controls are placement and replacement. An XWEB EVO that answers on the public internet is now a known target with public technical detail and working exploitation. Removing it from direct internet access, placing it behind a VPN or an allow-list, and monitoring for requests to the affected CGI paths are interim measures. They are not a fix.
The credential design deserves attention beyond this product. A password derived from the MAC address and the date is not a secret, because the MAC is not a secret and the date is public. Any vendor scheme that generates credentials from device attributes should be treated as a default password, not as authentication. The same reasoning applies to hardcoded recovery accounts and shared service accounts across a fleet.
The speed from publication to probing is the other lesson. Automated scanners watch CVE feeds and reach exposed hosts within hours, so the window between disclosure and exploitation is effectively zero for anything reachable from the internet. Operators should assume that an exposed OT device is found on the day its vulnerability becomes public, and should know which of their devices are in that position before the advisory lands.
Sources: SecurityWeek: Hackers probe newly disclosed flaws in Copeland refrigeration controllers; VulnCheck: Code of the day, unauthenticated root in Copeland XWEB EVO.
3. A drone maker’s update channel is the suspected path into QBase
Source and event dates: Quantum Systems confirmed the incident in a statement reported on November 16, 2026, and said it occurred at the end of October 2026. Der Spiegel and the Financial Times reported the suspected malicious update, and Security Affairs summarized that reporting on November 16. The company statement and the reported details are separate, and the reported details remain unconfirmed by the company.
News briefing
Quantum Systems, a German defense company that builds reconnaissance and interceptor drones and the software that controls them, confirmed that it identified a cybersecurity incident at the end of October. The company said an internal investigation found an active attack by hostile actors, that the threat was contained with the help of its IT and security teams and government authorities, and that there is no risk in operating its unmanned systems for end users and operators. It reported the incident to authorities, and an investigation is continuing.
Der Spiegel and the Financial Times, citing internal sources, reported that attackers may have used a malicious update file for QBase, the company’s Windows software for controlling drones and planning missions. That would imply compromise of an update server rather than of the drones themselves. An unexpected software update reached customers and raised alarms inside the company, according to those reports, and it is not clear how many customers installed it.
The reported consequences reached military operations. Der Spiegel reported that the German military grounded Quantum Systems drones as a precaution and excluded them from a recent exercise. Military counterintelligence, the domestic intelligence service and the federal cybersecurity agency are reportedly investigating. None of those details has been officially confirmed, and no attacker has been identified. The company said European defense companies, particularly those supporting Ukraine, have faced persistent attacks for years.
OT implications
The software distribution channel is part of the attack surface. If an update server is compromised, the trust relationship between vendor and customer becomes the delivery mechanism. Signing updates is necessary but not sufficient if the signing keys sit on the same infrastructure that distributes them, or if the endpoint does not verify the signature before applying the package. Build systems, signing services and distribution servers should be separated, and customers should be able to verify an update independently.
Staged rollout and rollback are the operational controls that limit the damage. A canary group, a defined pause window, and the ability to identify which devices received which version turn a malicious update into a contained incident rather than a fleet-wide event. That requires an inventory that records software versions per device, not a spreadsheet of purchase orders.
Asset owners should ask the same questions of every OT vendor whose tools update controllers, HMIs and engineering workstations. How is the update signed, where are the keys held, does the device verify the signature, and how quickly can the vendor and the customer determine what was installed? The answers belong in procurement and in the incident response plan, because the update path is often the most trusted path into an operational network.
Sources: Security Affairs: German military grounds Quantum Systems drones following reported cyberattack. The report cites Der Spiegel and the Financial Times for the suspected malicious update.
4. Ransomware set an October record, and manufacturing ranked third
Source and event dates: Check Point Research published its October 2026 Cyber Threat Landscape in the week of November 9, 2026. Industrial Cyber reported the figures on November 16. All data refer to October 2026.
News briefing
Check Point Research recorded 1,063 ransomware attacks in October, a record and a 35% increase over the same month last year. Organizations averaged 2,550 attacks per week, down 9% from September but still up 32% year over year. The monthly easing is real, and so is the annual trend.
Industrial manufacturing accounted for 11.6% of reported ransomware victims, third behind business services at 31.5% and consumer goods and services at 14.9%. Energy and utilities accounted for 4.1%. Those shares sit alongside much larger attack volumes: energy and utilities averaged 2,972 attacks per week, up 21% year over year, and industrial manufacturing averaged 1,838, up 20%. Education remained the most targeted sector at 5,934 attacks per week, up 33%, while agriculture recorded the sharpest rise at 93%, to 3,218. Telecommunications averaged 3,068 attacks, up 19%, hospitality 3,053, up 46%, and hardware and semiconductors 2,550, up 74%.
Europe recorded the fastest regional growth at 39% year over year. Africa had the highest volume at 3,834 weekly attacks per organization, followed by Latin America at 3,483 and Asia-Pacific at 3,436. North America accounted for 48% of reported ransomware incidents, Europe 22% and Asia-Pacific 18%.
The report also tracked a new extortion group. n0n appeared in September 2026, claimed more than 170 victims by early November, and took 8% of published attacks in October, entering the top three within a month. Its operators leaned on AI-assisted tooling, but the observed initial access came through weak or stolen credentials rather than exploited vulnerabilities. The Gentlemen led with 13% of published attacks and also operates as an initial access broker, while Qilin followed at 12%. Check Point reported that one in 35 enterprise generative AI prompts carried a high risk of sensitive data leakage in October, affecting 90% of organizations that regularly use such tools.
OT implications
The manufacturing and energy numbers matter because both sectors sit close to operational technology. A ransomware event that starts in business services or on a plant-adjacent server can still stop a line, and the annual trend is upward even when a single month dips. The sector share of victims is not the same as the sector share of impact, because a small number of manufacturing incidents can carry disproportionate downtime.
n0n’s credential path is the detail to take into an OT program. Stolen VPN, remote monitoring and management, and shared administrator credentials are common routes from IT into OT, and AI-assisted tooling shortens the time between access and extortion. Multifactor authentication on every remote path, unique credentials per device, and prompt removal of accounts for departed staff and contractors reduce the value of a stolen password. Segmentation limits what a stolen credential can reach.
For OT specifically, recovery is the constraint. Offline backups of controller configurations and engineering workstations, tested restore procedures, and a documented order of restoration matter more than detection alone. An asset inventory is the prerequisite for all three, because a team cannot restore or isolate devices it has not identified.
Sources: Industrial Cyber: Check Point finds ransomware at record levels as n0n rises into top three; iTWire: Cyber attacks remain a third higher than last year as AI-powered extortion group breaks into the top three.
5. Offshore energy told to engineer connectivity resilience, not bolt it on
Source and event dates: Marlink published the analysis in its Knowledge Hub in November 2026, and Industrial Cyber reported it on November 16. Marlink launched its OT Security offering in late August 2026.
News briefing
Marlink published an analysis arguing that network resilience is a critical requirement for energy operations rather than an availability nicety. Robert Austen, vice president of energy for the United Kingdom and Middle East at Marlink EEG, wrote that connectivity disruptions in offshore environments can interrupt production, compromise safety monitoring, and deprive remote assets of the real-time visibility needed to respond to emerging risks. Automation, remote monitoring and data-driven decision-making have made operations more dependent on that connectivity, not less.
Austen described a multi-orbit approach that combines low Earth orbit, geostationary and medium Earth orbit satellite services with terrestrial and 4G/5G networks, so that operations are not dependent on a single network. He paired that with intelligent orchestration that monitors connectivity, routes traffic and prioritizes critical applications such as SCADA and remote monitoring. On the security side, he pointed to network segmentation and passive monitoring as the means to separate operational systems from corporate networks and to detect threats without disrupting processes.
“Resilience is not a simple failover plan added to the network after the fact,” Austen wrote. “It must be built in from the start.” He argued that access to multiple networks is only part of the solution and that management and security have to be part of the same design. Marlink itself launched an OT Security offering for maritime and land-based industries in late August, combining professional services with OT-native technology.
OT implications
Remote and offshore sites often have one path to the internet and one path to management. That concentration is a single point of failure for operations and for security monitoring at the same time. When the link drops, the plant keeps running but the security team goes blind, and the visibility gap can outlast the outage because logs and alerts are lost.
Resilience therefore needs local autonomy. Controllers should keep running with local control and annunciation, telemetry should store and forward, and security policy should be enforced at the edge rather than only in a central console. Passive monitoring is the right default for OT, but it has to be paired with an inventory and with a way to act on what it finds, including when the link is intermittent.
Failover should also be tested with the security stack in the path. Many designs restore connectivity by routing around the inspection point, which solves the availability problem and removes the visibility that the architecture was built to provide. The test is whether the site can be operated, monitored and patched through the backup path, not simply whether it can reach the internet.