Anthropic launches effort to secure critical infrastructure

Anthropic’s announcement page for the Cyber Mission is dated Oct. 8, 2026.[1] Anthropic frames the effort as a long-term attempt to secure systems people depend on, including power grids, water systems, transportation networks and government systems.[1] The program is designed to bring frontier Claude models, on-site engineers and threat research to providers defending operational technology.[1] Anthropic says it will begin with a small group of providers, using the initial work to learn which approaches are effective and practical.[1] Companies that build critical-infrastructure security products or services can register their interest.[1]

Power grids, water utilities, factories and transportation networks run on controllers, control software and industrial networks, Anthropic says.[1] Anthropic says the program’s resources are intended to help participating providers identify and fix vulnerabilities in their customers’ systems.[2] For infrastructure operators, the stated focus is therefore not only finding weaknesses, but addressing them in environments where maintenance can be constrained by the need to keep systems running.[1][2]

The company says several partners are already working with Claude to fix vulnerabilities and help their customers do the same.[1] Rockwell Automation said responsible, rigorously validated AI could strengthen cybersecurity across critical infrastructure, while PwC’s OT security lead called for coordinated, industry-wide work to identify vulnerabilities and mitigate risks.[1] Anthropic also said Project Glasswing partners found many vulnerabilities but had not yet reduced cyber risk enough.[1] That statement signals a remaining challenge, rather than evidence that the new effort has already lowered risk.

Anthropic has not specified whether partners will receive free access to its models or who will pay the associated AI computing costs, Axios reported.[2] Axios also reported that it was unclear how participants could test and deploy fixes without disrupting utility operations.[2] Those unanswered questions matter to providers weighing the program: the offer includes models, engineers and research, but the available evidence does not establish the commercial terms or a deployment process.[1][2] Anthropic says its initial small-group approach is intended to help identify practical methods.[1]