← News
/DISPATCH · 2026-08-21

CISA/NSA/FBI/DOE/EPA Joint Advisory: AI-Generated Scripts Actively Attacking Siemens S7 PLCs Have Entered Pre-Positioning Across Multiple Critical Infrastructure Sectors

NSA, CISA, FBI, DOE, and EPA jointly issued advisory AA26-231A on August 19, confirming attackers are using AI-generated Python scripts to actively scan and probe Siemens S7-series PLCs. Targets span energy, water, chemicals, critical manufacturing, and the Defense Industrial Base. Block TCP/102 at the perimeter immediately and inventory engineering workstations.

NSA, CISA, FBI, DOE, and EPA jointly issued security advisory AA26-231A on August 19, 2026, confirming attackers are actively scanning and probing Siemens S7-series programmable logic controllers (PLCs) with AI-generated Python scripts. Coverage includes S7-200, S7-300, S7-400, S7-1200, S7-1500, and the full F-series safety PLC family. The advisory states this is a live, ongoing threat affecting critical manufacturing, energy, water and wastewater, chemicals, food and agriculture, and commercial facilities. The Defense Industrial Base is named because it likewise depends on S7. Inventory engineering workstations immediately and block TCP/102.

Incident Overview

AA26-231A is jointly signed by the five agencies. Current attacks are in the pre-positioning phase. Attackers have not yet issued destructive write commands at scale, but are using internet scanning services such as Censys and ZoomEye to identify exposed or weakly segmented S7 PLCs and performing read operations to learn the target environment. Key risk at this stage: once read access is obtained, attackers can map process flows, identify control logic, and understand safety dependencies, preparing for later writes (modify logic, setpoints, alarms, safety interlocks, emergency shutdown).

The advisory notes threat actors are using legitimate open-source industrial automation tools snap7.dll and the python-snap7 library—standard components for system integrators and automation engineers. Attackers wrap snap7 calls with AI-generated packaging that resembles legitimate OT monitoring tools, evading security-team attention. AI lets attackers build custom scripts without deep ICS protocol expertise. ICS attack bar drops significantly.

Technical Details

Attack path has two stages:

AI's role in this campaign:

The read phase yields PLC project files, user programs, HMI tags, and safety-system configuration—enough to draw process flow diagrams and identify critical safety interlocks. Destructive writes not yet issued; physical-damage capability already present.

Impact Assessment

Siemens S7 is a core node in global industrial control systems: power plants, substations, water treatment, chemical plants, food processing, assembly lines, commercial-building HVAC, and defense equipment. AA26-231A lists high-exposure sectors: critical manufacturing, energy, water, chemicals, food and agriculture, commercial facilities, and the Defense Industrial Base. Escalation from read to write spreads across process: process interruption (altered recipes or timing halt production), equipment damage (process parameters drive physical equipment into hazardous ranges), safety incidents (disabled interlocks or emergency shutdown systems create real personnel risk), data theft (recipes, process parameters, customer data), cascading impact (single-site outage propagates through the supply chain).

AA26-231A also flags third-party exposure: asset owners whose PLCs are remotely accessed by system integrators or managed service providers may not realize those PLCs are internet-reachable. This advisory follows a July compromise of 30+ Minnesota water-utility PLCs, further confirming that U.S. critical-infrastructure PLC exposure is under continuous scan.

Remediation

AA26-231A's action list is operational:

AA26-231A is the first federal advisory to explicitly list AI-assisted script generation as an active threat to critical infrastructure. OT security operations need real-time detection and automated response comparable to an IT SOC, plus dedicated protocol parsing at the OT/IT boundary.

References