When AI Lowers the Bar for Industrial Attacks: Defending OT in the Age of Generated Exploits

Five U.S. federal agencies — the NSA, CISA, the FBI, the Department of Energy, and the EPA — issued a joint advisory (AA26-231A) warning that threat actors are using AI-generated code to build exploitation scripts against Siemens S7 series PLCs at water, energy, chemical, and manufacturing sites. The advisory describes a four-stage attack chain in which attackers scan for controllers with port 102 exposed and use AI tools to generate working Python scripts via the open-source snap7 library, without describing any new vulnerability. Siemens confirmed the alert points to configuration weaknesses and exposure rather than a product flaw, and the advisory flags unauthorized snap7.dll use outside approved systems as a concrete indicator of compromise.

For years, the security of industrial control systems rested on an implicit barrier: attacking a programmable logic controller PLC required deep, specialized knowledge of proprietary protocols and hardware. In August 2026, five U.S. federal agencies issued a joint advisory confirming that this barrier is eroding. Threat actors are using AI-generated code to build exploitation scripts against Siemens S7 series PLCs deployed at water, energy, chemical, and manufacturing sites. The advisory is a milestone because it is the first government alert to state plainly that AI is being used to target operational technology OT . OT systems control physical processes. A compromised PLC can mean a halted production line, a tampered process parameter, or in the worst case a safety incident. If AI collapses the expertise required to attack these systems, then the population of capable attackers grows, and the old assumption that OT is protected by obscurity no longer holds. The advisory AA26-231A , co-signed by the NSA, CISA, the FBI, the Department of Energy, and the EPA, describes a specific technique. Attackers use internet scanning services such as Censys and ZoomEye to find Siemens S7 PLCs with port 102 exposed, then feed that targeting data into AI tools that generate working Python exploitation scripts using the open-source snap7 and python-snap7 libraries. Those scripts communicate over the S7comm protocol, which historically lacked authentication on many devices. The advisory covers the S7-200, S7-300, S7-400, S7-1200, and S7-1500 families. Critically, it does not describe a single new vulnerability. Siemens confirmed that the advisory points to configuration weaknesses and exposure, not a new product flaw. That distinction shapes the entire defensive response. The attack chain has four stages. First, discovery: scanning for exposed controllers on port 102. Second, generation: using AI to produce scripts that speak S7comm without the attacker needing to understand the protocol internals. Third, disguise: presenting the scripts as legitimate OT monitoring software so they blend into normal engineering traffic. Fourth, access: reading and potentially writing PLC memory, configuration data, and ladder logic. The reason this is hard to detect is that the traffic looks like ordinary engineering activity. There is no malware signature, because the tool is a legitimate library being used for an illegitimate purpose. The advisory names one concrete signal: unauthorized use of snap7.dll outside approved systems. A facility whose SCADA platform does not use snap7 that observes snap7 traffic on port 102 should treat it as an indicator of compromise. The response is about reducing exposure and controlling access, not waiting for a patch that may never come.