Advanced AI models can now successfully blueprint cyberattacks on critical infrastructure like power grids and water systems, according to new research from Frenos. This development poses a significant and urgent threat to operators of these essential services. Researchers tested ten AI models' ability to plan attacks against industrial control systems. Three of these models demonstrated alarming success rates in constructing attack plans across numerous attempts. While performance varied, even frontier models showed a concerning capability. The research involved thousands of tests where AI models were tasked with identifying attack steps based on target information. Critical services have always been targets for adversaries, but AI accelerates the threat. Many infrastructure operators struggle with resources and are hesitant to disrupt services for system upgrades. Industrial systems are particularly sensitive to even minor digital disruptions. AI can assist attackers by quietly mapping networks and identifying vulnerabilities tirelessly. AI can conduct reconnaissance and data analysis without triggering immediate security alerts. The research setup was relatively simple, suggesting that more advanced tools could further enhance AI's attack planning capabilities. Initiatives are emerging to provide resources and support for critical infrastructure organizations to address these growing AI-driven threats.