A joint federal advisory confirms adversaries are using AI-generated exploit code against industrial control systems, marking the first operationally confirmed use of AI-assisted intrusion tooling against US critical infrastructure.
Intelligence Lead
CISA, the NSA, the FBI, the Department of Energy, and the EPA issued a joint advisory (AA26-231A) on 19 August confirming that threat actors are using artificial intelligence to generate custom exploitation scripts against Siemens S7 Series programmable logic controllers, with confirmed intrusions at water and wastewater utilities across at least twelve states since 27 July. The advisory landed within days of a separate, reportedly Iran-linked cyberattack that took a small UK power generation facility offline for four days, indicating parallel probing of Western industrial control systems by state-linked actors.
Situation Report
The federal advisory describes threat actors using internet scanning services to identify exposed Siemens S7 PLCs running outdated firmware, then deploying AI-generated Python scripts built on the python-snap7 library and associated snap7.dll bindings to communicate directly with device-level controllers. Investigators assess the scripts were disguised as legitimate monitoring software, allowing operators to probe and manipulate industrial processes without triggering conventional intrusion-detection signatures tuned to known malware families.
Confirmed intrusions span water and wastewater treatment facilities in a dozen states, with the FBI and EPA noting the campaign has been active since at least 27 July. Sector exposure extends beyond water: CISA lists Critical Manufacturing, Energy, Chemical, Food and Agriculture, and Commercial Facilities as also within the targeting envelope, reflecting the generic reach of the S7 protocol across industrial environments rather than a campaign built narrowly for one vertical.
Separately, a UK power generation site, deliberately unnamed by authorities, was taken offline for four days in an intrusion attributed by US and UK sources to actors linked to Iran's Islamic Revolutionary Guard Corps. The facility was small and the outage did not affect the wider national grid, but investigators assess the operation's value to Tehran was demonstrative rather than disruptive: proof that IRGC-linked operators can reach and shut down Western energy infrastructure at will.
No attribution has been assigned to the Siemens S7 campaign at the nation-state level, and the joint advisory stops short of naming a sponsor. The proximity in timing and shared targeting logic across industrial control systems, AI-accelerated tooling, and parallel probing of US and UK infrastructure is, at this stage, a pattern rather than a confirmed link.
Background & Context
Siemens S7 controllers are among the most widely deployed programmable logic controllers in Western industrial environments, running processes from water treatment dosing to energy generation sequencing. Their ubiquity, combined with legacy deployments that predate modern network segmentation practice, has made them a recurring target since Stuxnet first demonstrated PLC-layer sabotage against Iranian centrifuges in 2010. The current campaign inverts that history: Iran-linked actors are now among the operators probing Western equivalents of the systems Iran itself once had disrupted.
The use of AI to generate working exploitation code against operational technology protocols has been anticipated by threat intelligence analysts for several years, but this advisory represents one of the first instances where government agencies have confirmed AI-assisted tooling in live intrusions against US critical infrastructure, rather than in red-team demonstrations or theoretical assessments.
Analysis & Assessment
The advisory's significance lies less in the immediate physical impact, which appears to have been contained, than in what it confirms about the trajectory of AI-enabled offensive cyber capability. Generative tooling lowers the technical barrier for producing working OT-layer exploits, which historically required specialist knowledge of proprietary industrial protocols. That barrier reduction is likely to expand the population of actors capable of credible infrastructure-targeting operations beyond the small set of states and criminal groups that have traditionally possessed it.
The parallel UK power plant incident is best read as a signaling operation. A four-day outage at an unnamed, minor facility generates limited disruption but considerable strategic communication: it tells Western governments that Iran-linked operators retain reach into national energy infrastructure despite sustained sanctions and cyber-defensive investment since the 2026 Iran war. Assessed with moderate confidence, this incident and the Siemens S7 campaign are unlikely to be centrally coordinated, but both reflect a shared judgment among state-linked actors that AI-assisted OT intrusion has crossed from theoretical to operational.
Watch for follow-on advisories naming additional PLC families or industrial protocols as targets, and for confirmation or denial of Iranian state sponsorship of the UK incident from British authorities in the coming weeks.