The Canadian Centre for Cyber Security has issued an advisory warning Canadian organizations about an active global campaign against programmable logic controllers (PLCs) that are exposed to the internet. The new element in this advisory is the method: the Cyber Centre says threat actors are using artificial intelligence to find these devices and attack them.

This isn't a fresh warning about industrial control systems in general. The Cyber Centre and its partners have previously flagged attacks on internet accessible industrial control systems and operational technology (ICS/OT) aimed at disrupting water facilities and other critical services in Canada and abroad. What has changed is the tooling. AI makes the discovery and targeting phase faster, which means the window between "we put that PLC on a public IP" and "someone found it" keeps shrinking.

The advisory points to concrete incidents. In October 2025, the Cyber Centre observed several incidents affecting ICS/OT systems in Canada. In one of them, a threat actor manipulated water pressure valve controls at a municipal water system, causing what the Cyber Centre describes as minor impacts to water service delivery. Minor is the word the regulator chose. It's also the word you get when someone else's hand is on your valves and you were lucky.

The advisory also cites reporting from the United States. According to the Cyber Centre, the Cybersecurity and Infrastructure Security Agency (CISA) reported a significant increase in attacks against PLCs at water and wastewater utilities. In those cases, threat actors changed passwords to lock operators out of their own controllers and took devices offline by changing their network addresses. The result was boil water advisories and sustained manual operations at affected utilities. Operators had to run plants by hand because someone on the internet held the admin credential they couldn't get back.

Read that sequence again with an identity hat on. The attackers didn't exploit exotic firmware bugs, at least not according to what the Cyber Centre describes. They found a device on the open internet, authenticated to it, and changed the password. That's an access control failure first and an OT failure second. A PLC reachable from any network with a default or shared credential isn't a hardened industrial asset, it's a web app with a really expensive backend.

The advisory does not, in the material released, name affected organizations, list specific PLC vendors or models, or describe the AI tooling in detail. It also doesn't attribute the campaign to a particular actor. Those gaps matter: treat any claims about which brands are being hit or which models are being used to hit them as unverified until a primary source says otherwise.

Who is affected? Anyone running ICS/OT with a management interface reachable from the internet, and the advisory calls out water and wastewater specifically. But the pattern (find exposed device, log in, lock out the owner) works on any controller with a network port and a login prompt. Municipal utilities are the headline because the consequences are public and measurable. Manufacturing, building automation and energy operators shouldn't assume they're too boring to be scanned. Scanning at scale is exactly what automation is good at, and nobody's PLC is too boring for a script.

The full advisory is available from the Cyber Centre at cyber.gc.ca.

The encouraging part is that the fixes here are old and they work. Identity and security teams should start by getting an honest inventory of every OT device with an interface reachable from outside the plant network, then remove that exposure or put it behind a VPN or jump host that requires MFA and named accounts. Check whether PLC and HMI credentials are default, shared across sites, or unchanged since commissioning, and rotate them into a vault with an owner and a recovery path so a hostile password change doesn't lock you out for good. Finally, confirm you're logging authentication events and configuration changes on those devices and that someone actually reviews them, because a password change at 2 a.m. from an unfamiliar address is only useful evidence if a human sees it before the boil water notice goes out.