Anthropic is offering frontier models, engineering support and threat research to infrastructure defenders, while launching free recurring security scans for enrolled open-source projects.
Anthropic announced the Anthropic Cyber Mission on October 8, introducing two initiatives aimed at helping defenders secure critical infrastructure and open-source software. The company says its program for critical infrastructure will provide frontier Claude models, on-site engineers and threat research to organizations that protect operational technology—the control systems used by power grids, water utilities, factories and transportation networks. Anthropic says these systems can be difficult to patch because they often cannot be taken offline and may depend on proprietary equipment.
The first cohort includes 11 companies: Accenture, Booz Allen, CrowdStrike, Deloitte, Dragos, Hitachi, Insane Cyber, Nozomi Networks, Palo Alto Networks, PwC and Rockwell Automation. Anthropic says some partners are already working with Claude to address vulnerabilities and help customers do the same. The company describes the initial group as a way to learn which approaches are practical; it has not announced broad availability to infrastructure operators or quantified security outcomes.
The second initiative, OSS Scanner, offers enrolled open-source projects periodic scans from Anthropic’s most capable models at no charge. Reports can include a proof of concept, an explanation of a possible vulnerability and a suggested fix. Unlike human-verified disclosures, these reports are generated and sent without human review. Anthropic says it expects a true-positive rate above 90%, but warns that reports can contain inaccuracies, including incorrect severity ratings. The company says projects that cannot manage unreviewed reports will continue to receive human-verified disclosures through its coordinated disclosure process.
The launch comes as AI tools are being positioned both as a way to accelerate vulnerability discovery and as a potential aid to attackers. Anthropic’s plan puts model access alongside security specialists and established vendors rather than presenting the models as a standalone fix. Its own announcement acknowledges that AI cannot address every infrastructure risk. Whether the programs lead to faster remediation will depend on operators’ ability to validate findings, prioritize them and safely apply changes to systems that may be difficult to interrupt.