AI-Generated Exploits Target Siemens S7 PLCs Across Critical Infrastructure
Five US federal agencies — NSA, CISA, FBI, Department of Energy, and EPA — issued joint advisory AA26-231A on August 20 warning that threat actors are using AI-generated exploitation scripts to attack Siemens S7 Series programmable logic controllers across energy, water and wastewater, chemical, food and agriculture, and critical manufacturing sectors. The advisory describes the campaign as an "active threat" — not theoretical, not reconnaissance-only, but ongoing exploitation of internet-exposed PLCs.
The attack chain is straightforward and marks a genuine inflection point in ICS threats. Actors use internet scanning services (Censys, ZoomEye) to find exposed S7-200, S7-300, S7-400, S7-1200, and S7-1500 PLCs running outdated firmware or with weak configurations. They then feed that targeting data into AI tools that generate functional Python exploitation scripts using the open-source snap7.dll and python-snap7 libraries — the same libraries legitimate engineers use for PLC monitoring and programming. The scripts are disguised as legitimate monitoring tools and provide read/write access to PLC memory, configuration data, and ladder logic via the S7comm protocol.
This is the first confirmed joint-agency warning that names AI-generated exploit code as an active component of a critical infrastructure attack campaign. The scripts are not proof-of-concept or researcher tools; they are functional weaponised code being deployed against production systems that control water treatment, power generation, and chemical processing. The advisory explicitly states that exploitation could lead to disruption of industrial processes, safety incidents, equipment damage, and data compromise.
The targeted S7 Series PLCs span every major generation Siemens has shipped. The S7-1500 F-series (failsafe) controllers are specifically called out, meaning safety-instrumented systems are in scope. Many of these devices run end-of-life firmware or use legacy protocols like Modbus that lack authentication or encryption. Siemens has stated there are no new vulnerabilities in its products — the attack exploits known weaknesses and misconfigurations identified in a July 2025 advisory (SSA-104599).
The broader context matters. This campaign follows a wave of Iran-nexus attacks against water and wastewater facilities across at least 12 US states, where operators were temporarily locked out of their own PLCs. Whether the current campaign originates from the same actors or from opportunistic copycats exploiting the same exposed attack surface is unclear. The advisory does not attribute the activity to a specific nation-state group.
Separately, the same week saw reporting on a near-autonomous multi-agent AI attack framework (Hermes/OpenClaw) that targeted Taiwanese government infrastructure over 3-4 July, cracking 85 credentials, exfiltrating 2,564 personnel records, and establishing persistent backdoors across multiple government portals. That campaign produced 1,395 files of output and expanded to target IT supply chain vendors, a nuclear safety agency, and seven energy sector companies. The cost of running a competent attack has collapsed. The cost of defending against one has not.
So What / Action
If you operate Siemens S7 PLCs in any critical infrastructure sector, this advisory demands immediate action. Audit external-facing PLCs for internet exposure — if any S7 controller is reachable from outside your OT network, take it offline or isolate it now. Ensure firmware is current on all S7 Series devices, particularly the S7-1200 and S7-1500 families which remain widely deployed. Enable multi-factor authentication wherever the management interface supports it. Deploy ICS-aware monitoring to detect anomalous S7comm traffic, unexpected connection attempts from snap7-based tools, and credential access patterns that don't match legitimate maintenance windows. The barrier to ICS exploitation has just been lowered by an order of magnitude; your defensive baseline needs to move up by the same margin.
Jonathan Care has 33 years in cybersecurity and fraud detection. These are his personal views, not those of his employer.