US agencies warn hackers are using AI to break into Siemens factory controllers

Key points

• Five US agencies, led by CISA, warned on the active targeting of Siemens S7 Series programmable logic controllers used across manufacturing, energy, water, chemicals, food, and commercial facilities.
• Attackers are using AI to write Python scripts that mimic legitimate monitoring software and talk to the controllers over the S7comm protocol.
• All S7 model families are in scope: S7-200, S7-300, S7-400, S7-1200, and S7-1500, including the F-series safety controllers.
• The hackers find targets through public internet-scanning services such as Censys and ZoomEye, then log in using default or weak passwords.
• Agencies assess the activity is reconnaissance and pre-positioning, meaning the groundwork for future attacks that could disrupt physical processes.

Five US federal agencies have issued a joint warning that hackers are actively going after Siemens factory-floor computers, and using artificial intelligence to speed up the work.

The advisory comes from the National Security Agency, the Cybersecurity and Infrastructure Security Agency (CISA), the FBI, the Department of Energy, and the Environmental Protection Agency. It focuses on Siemens S7 Series programmable logic controllers, or PLCs, which are the small industrial computers that open valves, run motors, and keep production lines moving inside factories, power plants, and water utilities.

The agencies say this is not a drill. Attackers are already inside the reconnaissance phase, mapping US installations and testing what they can do.

Which devices are being targeted?

Every current Siemens S7 family is on the list. That covers older S7-200, S7-300, and S7-400 units still running in long-life plants, plus the newer S7-1200 and S7-1500 lines, including the F-series safety controllers that trigger emergency shutdowns.

The sectors most in the crosshairs are critical manufacturing, energy, water and wastewater, chemicals, food and agriculture, and commercial facilities. Siemens PLCs also sit inside defence industrial base plants, which the agencies say could be hit next.

How are the hackers getting in?

They are looking for controllers plugged straight into the public internet. Once found, they log in with default or barely-changed passwords, then run scripts that speak the controller's native language.

To find targets, the attackers use commercial internet-scanning services like Censys and ZoomEye, which index every device visible online. That gives them a shopping list of exposed PLCs running outdated firmware.

The scripts themselves are the interesting part. The CISA advisory describes AI-generated Python code built around snap7 and python-snap7, open-source libraries that talk to Siemens controllers over the S7comm protocol, the standard way S7 devices exchange data. Wrapped up, the tools look like ordinary factory monitoring software, which is what lets them slip past security teams.

With that access the attackers can read and write to the PLC's memory, its configuration, and its ladder logic, which is the program that tells the machine what to do.

Should operators actually worry?

Yes. The agencies say the read access seen so far looks like homework for future write operations, meaning the attackers are learning the plant so they can later change how it runs.

The potential damage list in the advisory is blunt: disrupted production, safety incidents if interlocks or emergency shutdowns are tampered with, physical damage to equipment run outside safe limits, stolen process recipes, and knock-on effects across connected suppliers.

Using AI to write the exploit code matters because it collapses the skills gap. Attacking industrial kit used to demand years of specialist knowledge. A large language model can now produce a working script from public vulnerability writeups in minutes.

What should operators do now?

CISA's top actions are practical. Inventory every S7 PLC, apply Siemens' security patches, and get the devices off the public internet. Tighten authentication, replace default credentials, and watch for unusual S7comm traffic or unexpected writes to data blocks.

Pay particular attention to third-party integrators and service providers with remote access. Asset owners often do not realise a contractor has punched a hole through the firewall for convenience.