Five US Agencies Say Attackers Are Writing PLC Exploits With AI
The NSA, CISA, FBI, EPA and DOE have issued a joint advisory on attacks against Siemens S7 controllers. The notable detail is that the exploitation scripts are AI-generated.
Five US federal agencies — the NSA, CISA, FBI, EPA and DOE — have issued a joint advisory about attackers targeting Siemens programmable logic controllers across critical infrastructure. The targeting is not the part to pay attention to. The tooling is.
The agencies say the attackers are using AI to generate their exploitation scripts.
The technical shape of it
Threat actors are scanning the internet for exposed PLCs and building exploits capable of disrupting industrial processes. The agencies list the potential impacts plainly: equipment damage, safety incidents affecting workers, compromise of sensitive data, and cascading effects across supply chains, connected facilities and business operations.
Targeted devices span the S7-200, S7-300, S7-400, S7-1200 and S7-1500 series — for most of them, regardless of CPU variant.
The sectors named are energy, critical manufacturing, water and wastewater, food and agriculture, chemical, and commercial facilities.
The method is the part practitioners should read twice. Open-source industrial automation libraries — snap7.dll and python-snap7 — are being combined with AI-generated scripts to produce tools that mimic legitimate OT monitoring software. Those tools let attackers tamper with PLC memory, configuration data and ladder logic programs.
A tool that looks like your monitoring software, speaking the protocol your monitoring software speaks, is a hard detection problem by construction.
What the agencies are claiming about AI
The advisory's own framing is that AI-generated exploitation scripts represent an evolution in capability, sharply reducing the technical expertise and time needed to build working ICS exploitation tools, and letting adversaries pick up new attack vectors and adapt to defences faster.
Note what that claim is and isn't. It is not that AI has invented a new class of attack against PLCs. It is that the specialist knowledge which previously gatekept ICS exploitation — protocol internals, ladder logic, vendor quirks — is no longer the barrier it was.
ICS security has quietly depended on that barrier for years. Obscurity was never a control, but it was a real cost imposed on attackers. That cost has dropped.
Reconnaissance, not impact — yet
The advisory describes an active threat rather than a theoretical risk, but it names no high-impact attacks observed in the wild. The agencies assess that the actors are conducting persistent reconnaissance in preparation for future disruptive or destructive operations.
The threat actors have not been attributed. The alert does follow a series of Iran-linked attacks on the US water sector: at least 12 states have seen attacks on OT systems, with no confirmed water supply disruptions. CISA had already urged the water sector to protect PLCs, and the US government previously warned about Iranian actors targeting controllers from Siemens, Schneider Electric and Rockwell Automation.
The recommendations, and the honest problem with them
The advisory asks organisations to patch, to isolate PLCs from the internet unless genuinely necessary, to enforce strong access controls, and to deploy security products that monitor ICS environments.
None of that is new, and that is precisely the issue. Every one of these controls has been standard guidance for a decade, and internet-exposed PLCs remain trivially findable. What has changed is the speed at which an exposed device becomes an exploited one.
If your OT asset inventory is out of date, that is now a shorter-fused problem than it was last year.
Source: SecurityWeek — Hackers Using AI to Target Siemens PLCs in Critical US Sectors