Teams that run power grids, water treatment and transit networks can now get top-tier AI plus engineers on-site to hunt and fix vulnerabilities.
Anthropic launched the Critical Infrastructure Defense Program as part of its Anthropic Cyber Mission. The program gives trusted providers access to its most capable Claude models, embeds on-site engineers, and shares threat research to help defenders secure the operational technology behind utilities, factories, transportation, and government systems. Eleven founding partners are named, including Accenture, Booz Allen, CrowdStrike, Deloitte, Dragos, Hitachi, Insane Cyber, Nozomi Networks, Palo Alto Networks, PwC and Rockwell Automation. Anthropic says several partners were already using Claude to find and remediate vulnerabilities.
This matters because it changes who can put frontier AI inside the systems that keep cities running. Until now, advanced models were usually available only through public APIs or kept inside a lab. This program creates a direct, vetted path for those models to be embedded into live defence workflows via trusted vendors.
Think of it as delivering a specialized toolkit plus an expert technician. The AI surfaces likely flaws and suggested fixes; the on-site engineers validate, test and apply them. Anthropic’s threat research helps partners decide which issues to address first.
Practically, operators should see faster detection and remediation, and security firms will move from demos to frontline deployments. It also raises clear governance questions about oversight, transparency and who gets access to these capabilities.
The open question now is whether these partnerships will produce auditable, repeatable results and governance that operators and regulators trust. Watch for operational rollouts, third-party audits, or competitors doing the same.
