Anthropic has launched the Anthropic Cyber Mission, a long-term initiative to help secure critical infrastructure and open-source software through artificial intelligence (AI) tools, security research, engineering support and funding.
The initiative begins with two focus areas: protecting operational technology (OT) used in essential services such as power grids, water systems and transportation networks, and identifying and fixing vulnerabilities in open-source software. Anthropic said the programme aims to strengthen cyber defences amid growing concerns about the potential misuse of frontier AI models for cyberattacks.
As part of the initiative, Anthropic has introduced the Critical Infrastructure Defense Program (CIDP), which will provide frontier Claude models, on-site engineering expertise and threat research to organisations responsible for securing industrial systems and government infrastructure.
The programme’s founding partners include Accenture, Booz Allen, CrowdStrike, Deloitte, Dragos, Hitachi, Insane Cyber, Nozomi Networks, Palo Alto Networks, PwC and Rockwell Automation. These organisations include cybersecurity providers, industrial equipment manufacturers and technology and consulting firms supporting critical infrastructure operators.
Anthropic said the programme is already underway, with several partners using Claude to identify and address vulnerabilities and help customers strengthen their systems. The initial phase will involve a small group of providers to assess which approaches are most effective and practical for infrastructure security.
The company highlighted the challenges of protecting OT environments, where equipment can remain operational for decades and cannot easily be taken offline for software updates. As a result, known vulnerabilities may remain unresolved for extended periods, while changes to industrial systems can carry operational risks.
Alongside CIDP, Anthropic has launched OSS Scanner, a free service that offers open-source projects regular security scans using its most capable models. The tool is intended to help maintainers identify vulnerabilities and develop patches for widely used software components.
The new efforts build on lessons from Project Glasswing, which uncovered vulnerabilities but also highlighted the difficulty of verifying findings, prioritising risks and implementing fixes. Anthropic said it plans to expand the Cyber Mission with additional tools, research and resources to help defenders secure systems and software.
