Anthropic Signs CrowdStrike, Hitachi, and 9 More to Defend Power and Water
Anthropic is launching the Anthropic Cyber Mission with two initiatives: the Critical Infrastructure Defense Program and OSS Scanner.
Critical Infrastructure Defense Program
Anthropic will provide its frontier Claude models, engineers, and threat research to companies guarding critical infrastructure, including power grids, water utilities, and transport networks. This program is a long-term effort to support defenders against cyber threats.
OSS Scanner
OSS Scanner is a free service that utilizes Anthropic’s strongest models to scan open-source code. It automatically sends bug reports directly to maintainers, eliminating the need for human review. Anthropic caps its liability at $1,000 per issue.
Founding Partners
Eleven companies have joined as founding partners:
- Accenture
- Booz Allen
- CrowdStrike
- Deloitte
- Dragos
- Hitachi
- Insane Cyber
- Nozomi Networks
- Palo Alto Networks
- PwC
- Rockwell Automation
These partners, including consultancies, security companies, and equipment makers, are already relied upon by operators in critical infrastructure sectors.
Details and Future Plans
Anthropic aims to initially work with a small group of providers to understand the most effective strategies. The program targets operational technology, which includes controllers, control software, and industrial networks that have been in use for decades.
Hitachi, in a separate statement, expressed its commitment to joining the program’s working groups and sharing its operational technology expertise.
Unanswered Questions
Anthropic did not disclose who will cover the computing costs or if partners gain free access to models. Axios reported that it's unclear how partners will test fixes without disrupting utilities.
Past Initiatives
In June, Anthropic launched a similar program for US state, local, tribal, and territorial governments, providing Claude models and technical support to over half of US states. In September, OpenAI committed $1 billion to cyber defense for water utilities and community banks.
OSS Scanner Performance
Over six months, Anthropic's models identified more than 29,000 candidate vulnerabilities. While staff could only review about 6,000, nearly 5,000 unchecked reports were sent to maintainers who requested them. Of the reports, Anthropic expects more than 90% to be valid.