Control Systems SecuritySubscribe

The Biggest OT and ICS Cyberattacks of 2026: What Caused Them and What Would Have Stopped Them

Seven incidents, the failure in plain words, and the control that would have mattered

Jeff Gray · · Control Systems Security

A 2026 wall calendar with dates circled in red hangs beside an industrial control panel showing an ALARM screen, with server racks and a dim plant tank in the background.

Illustrative image, not a photo of an affected site.

The biggest OT and ICS cyberattacks of 2026 were not clever. Almost every one started with a controller, HMI, or edge device that could be reached from the internet. Default or weak passwords showed up again and again, and so did a remote path nobody was watching.

The controls that would have stopped or limited them are the boring ones: Get control systems off the public internet, put every remote session through one managed and authenticated path, know what you own, and be ready to run by hand.

These are seven incidents from the year, each with the failure in plain words and the control that would have mattered.

1. Poland's energy sector wiper attack (December 29, 2025)

On December 29, 2025, coordinated destructive attacks hit more than 30 wind and solar farms, a combined heat and power plant serving almost half a million customers, and a manufacturing company. At the grid connection points, RTUs were damaged and the sites lost communication with the distribution system operator, although power production continued. At the heat and power plant, endpoint protection blocked the wiper before it could do its job.

Root cause: Public reporting points to vulnerable internet-facing edge devices for initial access, then default credentials to pivot onto HMIs and RTUs. At the plant, the attacker had been inside long enough to steal information and gain privileged accounts.

What would have stopped it: Retire or patch end-of-support edge devices. Change every default password, and write that requirement into integrator contracts. Plan for OT devices that come back bricked, because if you cannot verify the firmware, an attacker can destroy it. The plant already showed one control that worked: Endpoint protection on that network stopped the wiper.

2. The Iranian-affiliated PLC campaign (March to July 2026)

A joint advisory, AA26-097A, first published April 7 and updated July 22, describes Iranian-affiliated actors connecting directly to internet-exposed PLCs at water, energy, and local government sites. They pulled project files and modified logic, and they manipulated what operators saw on HMI and SCADA screens. The July update says the changes disabled shutdown and alarm logic, letting systems drift into unsafe conditions without telling operators.

The root cause was simple: PLCs answering on their native programming ports from the open internet, reachable with the vendor's own engineering software.

What would have stopped it: Put PLCs behind a firewall or secure gateway with no direct inbound access. On controllers that have a physical mode switch, leave it in Run so remote program changes are refused. Keep a known-good copy of every project file and compare against it. If your alarm logic can be rewritten from a rented server overseas, your alarm is a suggestion.

3. The summer of small water systems (July and August 2026)

What happened: On July 26 and 27, more than 30 Minnesota community water systems were targeted in what state officials called a coordinated cyberattack. In Braham, attackers shut down the operating controls for the well and treatment plant, and crews had it running again in about an hour and a half. Other cities switched to manual workarounds. In late August, two small private Colorado utilities, each serving fewer than 200 people, had equipment settings changed, remote access and alarms disabled, and pumping cycles altered. Water quality was reported safe in both states.

Officials have not published the full entry path for each site, and attribution was not confirmed. What is public matches a familiar setup. These were small utilities with remote monitoring and control equipment reachable from outside, and the attackers changed settings directly on that equipment.

What would have stopped it: Start with an honest inventory of every modem, PLC, and HMI that touches the internet, then remove that exposure. The cities that kept water flowing also leaned on contingency procedures and manual operations, so practice running by hand before a real outage forces it.

4. Ransomware stops a national dairy brand (July 2026)

Coca-Cola disclosed that a ransomware attack at its fairlife subsidiary reached systems including production-related systems, and production at fairlife's four U.S. facilities was temporarily suspended. By July 27, most production had resumed. The company said data was taken and that product safety was not affected.

The entry point has not been made public, so I will not guess. What we can say is the outcome: A ransomware event that touched production systems turned into a plant shutdown.

What would have stopped it, or at least shortened it: A real boundary between business IT and production networks, so an IT compromise does not automatically become an OT outage. Tested, offline backups of production servers and controller configurations. A recovery plan that names what gets restored first. Inventory on the shelf bought time here, and most plants do not have that kind of cushion.

5. France's supermarket refrigeration attack (October 2026)

At one French supermarket chain in October, refrigeration ran at 50 degrees Celsius for six hours and all the food was thrown out. Attackers had compromised remote refrigeration controllers from Danfoss, Copeland, and Dixell. France's finance minister told lawmakers about 700 of the country's roughly 10,000 supermarkets were affected.

Root cause: Management interfaces were exposed on public IP addresses, many of them on end-of-life controllers that no longer get security updates. Alert CERTFR-2026-ALE-012 said plainly that exposing a management interface on the internet is contrary to good practice, and warned that a factory reset is not enough if an end-of-life device stays exposed.

What would have stopped it: Take the controllers off the public internet. Give the refrigeration contractor access only through a secured VPN or protected network, change default passwords, and replace equipment the vendor stopped supporting years ago. Those public management interfaces were there for contractor convenience, and that is how the attackers got in.

6. Hacktivists and open VNC (advisory AA25-343A, December 2025)

Joint advisory AA25-343A warned that pro-Russia hacktivist groups were using minimally secured, internet-facing VNC connections to reach HMIs in water, food and agriculture, and energy. The advisory notes these groups often exaggerate their claims, but some intrusions caused real damage. The root cause was remote desktop tools left open to the internet with weak or no authentication, sitting in front of screens that can change a process.

What would have stopped it: Do not put VNC on the internet. Map every access point, put remote access behind strong authentication, and if a vendor needs to see an HMI, give them a managed, logged session that ends when the work does.

7. AI-assisted attacks on Siemens S7 PLCs (advisory AA26-231A, August 2026)

What happened: A joint advisory, AA26-231A, published August 19, describes an active campaign against Siemens S7 Series PLCs, the S7-200, S7-300, S7-400, S7-1200, and S7-1500 families, at sites in manufacturing, energy, water and wastewater, chemical, food and agriculture, and commercial facilities. Actors used internet scanning services to find exposed controllers. Then they used AI-generated scripts, built on public industrial libraries and disguised as legitimate monitoring tools, to read and write PLC memory, configuration, and ladder logic over the S7comm protocol. The advisory calls this persistent reconnaissance and capability development, preparing for later write operations that could disrupt a process. It does not name victims or confirm that this campaign already caused a plant outage. A Canadian government advisory on September 8 described the same activity as an active global campaign that was still underway.

Root cause: Controllers reachable from the internet, often with default or weak authentication and outdated firmware. AI adds speed, because it lets a less skilled actor turn public PLC documentation and open-source libraries into working exploit scripts in far less time.

What would have stopped it: Get the PLCs off the public internet, and block the engineering protocol (S7comm on TCP port 102) at the perimeter. Change default credentials, turn on controller password protection, and keep firmware current. Watch S7comm traffic for connections and writes outside maintenance windows, and treat unexpected monitoring tools on engineering workstations as hostile until proven otherwise.

The pattern

Look at all seven together and the same doors keep showing up. Nation-state crews, ransomware gangs, and hacktivists had different motives. They still used exposed controllers, unmanaged remote access, default credentials, and old edge devices nobody owned.

The useful part is that you can close those doors without buying a new platform. You need four habits:

  1. Know every device that touches your control network, and which ones the internet can see.
  2. Remove direct internet exposure, and route all remote access, including vendors, through one authenticated, monitored, time-limited path.
  3. Change default credentials and retire end-of-support devices on a schedule, not after an incident.
  4. Keep known-good logic and configuration backups, and practice running by hand.

By late 2026, attackers were still walking through the same exposed controllers, only faster. They used AI to write ICS exploit scripts, which cut the time and skill needed to hit those controllers, and they disguised the tools as ordinary monitoring software. The industry has to adapt now, rather than wait for the first confirmed plant-floor outage from this tradecraft. Run those four habits on a shorter cycle: Hunt internet-facing controllers continuously, not once a year, treat unscheduled engineering-protocol traffic and surprise monitoring tools as an incident, and practice manual operations before a scripted lockout forces the issue. AI made the old way in faster, and waiting for proof before acting is not a plan.

If remote access is where your site is weakest, I put the jump host, MFA, vendor workflow, and policy pieces into one working playbook: The OT Remote Access Playbook.


Sources

  1. Energy Sector Incident Report, 29 December 2025 (Jan 30, 2026)
  2. Alert on the Poland energy sector incident (Feb 10, 2026), edge devices and default credentials
  3. ESET Research, DynoWiper blocked by EDR
  4. Joint advisory AA26-097A, Iranian-affiliated PLC exploitation (Apr 7, 2026; updated Jul 22, 2026)
  5. Minnesota state release on the water system attacks (Jul 28, 2026)
  6. MPR News, 30+ Minnesota systems targeted (Jul 28, 2026)
  7. BleepingComputer, Minnesota water OT attack (Jul 29, 2026)
  8. SecurityWeek, Colorado water utilities (Sep 21, 2026)
  9. Reuters, Colorado water systems targeted in August (Sep 19, 2026)
  10. BleepingComputer, Coca-Cola 8-K on fairlife ransomware (Jul 16, 2026)
  11. fairlife / Coca-Cola release, production resumed (Jul 27, 2026)
  12. Alert CERTFR-2026-ALE-012 (Oct 12 to 16, 2026)
  13. SecurityWeek, ~700 stores, 50 C for six hours, EOL devices
  14. AP News, French supermarkets
  15. Joint advisory AA25-343A, pro-Russia hacktivists and VNC (Dec 9, 2025)
  16. Joint advisory AA26-231A, AI-assisted threat to Siemens S7 Series PLCs (Aug 19, 2026)
  17. BleepingComputer, U.S. warns of AI-powered attacks on Siemens PLCs (Aug 19, 2026)
  18. Canadian advisory, AI in an active global campaign against internet-exposed PLCs (Sep 8, 2026)

This article is an analysis of publicly available reporting, vendor and company statements, and published advisories only. No active scanning, probing, or unauthorized access was performed. No specific private facilities or operators are identified beyond what public reporting has already named.

OTFeed matches public ICS advisories to the gear a customer actually runs. otfeed.com

Analysis based solely on publicly available sources and open-source research. No active scanning, probing, or unauthorized access was performed. No specific private facilities or control systems are identified. The views expressed are the author's alone and do not represent the positions, strategies, or opinions of any government agency, department, or employer.