Mapping Critical Paths to Defend Against AI Attacks
Kristine Martz is Principal Product Advisor for Dragos Inc. and has over fifteen years of experience in power and utilities cybersecurity and regulatory compliance, with expertise in NERC standards and real-time systems security.
AI is changing how quickly adversaries can find their way around a network. The utilities that fare best will know, precisely and regularly, which paths lead to the systems that matter.
In early 2026, an intruder inside the enterprise IT network of a municipal water utility in Monterrey, Mexico, got help from an AI model that no one had asked to look for control systems. It found one anyway.
Dragos analysts who reviewed the evidence found that the model spotted an industrial gateway on the internal network, judged it a high-value target because of its proximity to operations, studied the vendor's documentation, and launched a large automated password attack against it. Every attempt failed, because none of the passwords it tried were valid.