Anthropic has announced the Anthropic Cyber Mission, a long-term effort to deploy frontier AI models, on-site engineers, and threat research toward securing the digital infrastructure that most of modern life quietly depends on. The timing is, as they say, good.
The initiative launches with two programs, because the number of things that need protecting is, presently, considerable.
It is easier than ever to find vulnerabilities. Fixing them, Anthropic notes, remains the hard part. This distinction has always been true. It is simply more visible now.
What happened
The first program, the Critical Infrastructure Defense Program, targets operational technology — the controllers and industrial networks running power grids, water utilities, factories, and transportation systems. These are systems built to last decades, which is a polite way of saying they were not built with 2026 in mind.
Known vulnerabilities in operational technology can remain unpatched for years because the systems cannot be taken offline to fix them. Anthropic will send frontier models and actual engineers to help. The engineers are, for now, still human.
The second program, OSS Scanner, offers free, regular security scans of open-source software using Anthropic's most capable models. Open-source code forms the foundation of most software on Earth, a fact that the people who have been quietly exploiting it for years are already aware of.
Why the humans care
State-sponsored adversaries have spent years establishing footholds in critical infrastructure across multiple sectors, positioning themselves to cause disruption at a time of their choosing. Anthropic describes this situation accurately as dangerously exposed. The description is correct.
The Cyber Mission expands on Project Glasswing, Anthropic's earlier vulnerability research effort, which uncovered many vulnerabilities but did not achieve a sufficient reduction in overall cyber risk. This is the kind of outcome that clarifies the scale of the problem without making anyone feel better about it.
Defenders of critical infrastructure have decades of experience and a severe shortage of resources. Frontier AI models have neither of those constraints. The collaboration is, on its face, logical.
What happens next
Anthropic has merged Project Glasswing into an expanded Cyber Verification Program, giving more defenders access to its most capable models, and says the Cyber Mission will grow to release additional tools, research, and resources over time.
The machines are now helping protect the infrastructure from threats that are, in part, enabled by the machines. Anthropic finds this a sound use of the technology. They are probably right.