Anthropic unveils programme to protect power and water networks – with 11 partners
Critical infrastructure defenders are to get access to Anthropic's most capable AI models and engineers. A free security scanner for open-source software comes on top.
Symbolic image · Photo: Novoklimov / Wikimedia Commons, CC BY 4.0 (cropped and resized)
Anthropic launched the “Anthropic Cyber Mission” on 8 October. The first part is a programme to protect critical infrastructure: security firms that guard power grids, water utilities and transport networks get access to Anthropic's most advanced Claude models, on-site engineers and threat research. The focus is on operational technology such as control systems in plants, substations and water facilities. The eleven founding partners include Accenture, Booz Allen Hamilton, CrowdStrike, Deloitte, Dragos, Hitachi, Insane Cyber, Nozomi Networks, Palo Alto Networks, PwC and Rockwell Automation. Anthropic did not disclose commercial terms.
The second element is the free “OSS Scanner,” which, on request, regularly scans open-source projects for vulnerabilities using Anthropic's strongest models. Reports go to maintainers without prior human review and may contain errors. In pre-launch testing, penetration testers checked 97 critical and high-severity findings from 48 projects; 85 met the bar for disclosure. Over six months, the models flagged more than 29,000 candidate vulnerabilities in widely used software, about 6,000 of which staff reviewed by hand.
Anthropic concedes that critical infrastructure is hard to defend in many ways that AI cannot fix. Finding vulnerabilities has never been easier, it says, but verifying and fixing them remains hard. The company forecasts that in two years AI will favour defenders – though in the near term that may not be true.