Anthropic launched the Critical Infrastructure Defense Program on October 8, 2026, as part of its Anthropic Cyber Mission to use AI safety expertise for protecting operational technology (OT) in energy, manufacturing, utilities and other critical sectors.
Anthropic said it will provide frontier Claude models, on‑site engineers and threat research to the partners to help identify and remediate vulnerabilities in OT systems before they can be exploited. Several partners are already using Claude to fix vulnerabilities and assist customers in doing the same, beginning with a small cohort to determine which approaches are most effective and practical.
The initiative focuses on operational technology such as power grids, water systems, factories and transportation networks, where equipment often cannot be taken offline for patching and known flaws may remain unaddressed for years. Anthropic warned that state‑sponsored adversaries have spent years establishing footholds in critical infrastructure and that the declining cost of exploiting vulnerabilities increases the need for defensive AI tools.
The company expects that within two years AI will advantage defense, making it easier to detect bugs before release and to actively protect systems with models, although currently verifying and fixing flaws remains slow and dependent on human effort. As part of the Anthropic Cyber Mission, Anthropic also released OSS Scanner, a free opt‑in service that provides regular vulnerability scans of eligible open‑source projects using its strongest Claude models, with an anticipated true‑positive rate above 90%.
OSS Scanner reports are generated without human review, which may contain inaccuracies such as incorrect severity ratings, while maintainers lacking capacity to triage raw findings continue to receive human‑verified disclosures through Anthropic’s coordinated vulnerability disclosure process. The Defender Advantage Fund, launched in August 2026, funds the free OSS Scanner and supports pilot programs for open‑source security groups including the Python Software Foundation, Apache Software Foundation, Alpha‑Omega and OpenSSF via the Linux Foundation.
Anthropic has previously funded the Python Software Foundation, Apache Software Foundation, Alpha‑Omega and OpenSSF through the Linux Foundation, and has supported Akrites and Gold Eagle to coordinate vulnerability reports. The Critical Infrastructure Defense Program draws on lessons from Project Glasswing, which was merged into an expanded Cyber Verification Program earlier in October 2026.
Since June 2026 Anthropic has supplied frontier Claude models and technical support to over 50 percent of U.S. states and several major public critical infrastructure operators through its earlier cyber defense program for state, local, tribal and territorial governments. Andrew Turner, president of commercial cyber at Booz Allen, called operational technology “the next frontier for autonomous AI-enabled attacks,” according to SiliconANGLE.
Join the discussion