Frontier AI in OT Defense: What's Real, What's Hype

OT security has moved through distinct eras defined by who was willing to attack it: decades of adversaries probing OT networks mostly for espionage rather than destruction, the first reusable ICS attack framework in 2021 (PIPEDREAM), militaries entering soon after, and state actors arming non-state proxies by 2024. We're now in the AI era, with adversaries using AI to accelerate OT attacks.

Dragos CEO and Co-Founder Robert M. Lee draws on frontline incident response, non-public work with frontier AI labs, and a real AI-assisted intrusion into a Mexican water utility to separate what's actually changing in OT threats from what's still hype. Distinguished Engineer Garrett Bladow then shows what Dragos found when it turned the same models on itself, using Anthropic's Project Glasswing to test frontier AI against the Dragos Platform.

What you'll learn:

• How OT threat activity has evolved era by era, and why AI marks the fastest shift yet
• What a real AI-assisted intrusion into a Mexican government water utility revealed, and what it does and doesn't prove
• Why AI-scale vulnerability deduplication, contextualization, and prioritization against operational context will outpace how most organizations rank risk today
• Why prevention-only strategies are just part of the equation, and what the SANS Five Critical Controls still get right
• How Dragos tested frontier AI against its own platform under Project Glasswing, and what that revealed about defending at scale